Faculty - Dashboard

@inproceedings{bib_Alig_2025, AUTHOR = {Srijith, Padakanti and Pahwa, Khushbu and Mamidi, Radhika and Surampudi, Bapi Raju and Gupta, Manish and REDDY, OOTA SUBBA }, TITLE = {Aligning Text/Speech Representations from Multimodal Models with MEG Brain Activity During Listening}, BOOKTITLE = {Conference on Empirical Methods in Natural Language Processing}. YEAR = {2025}}

Although speech language models are expected to align well with brain language processing during speech comprehension, recent studies have found that they fail to capture brain relevant semantics beyond low-level features. Surprisingly, text-based language models ex hibit stronger alignment with brain language regions, as they better capture brain-relevant semantics. However, no prior work has exam ined the alignment effectiveness of text/speech representations from multimodal models. This raises several key questions: Can speech em beddings from such multimodal models capture brain-relevant semantics through cross-modal interactions? Which modality can take advan tage of this synergistic multimodal understand ing to improve alignment with brain language processing? Can text/speech representations from such multimodal models outperform uni modal models? To address these questions, we systematically analyze multiple multimodal models, extracting both text- and speech-based representations to assess their alignment with MEGbrain recordings during naturalistic story listening. We find that text embeddings from both multimodal and unimodal models signif icantly outperform speech embeddings from these models. Specifically, multimodal text embeddings exhibit a peak around 200 ms, sug gesting that they benefit from speech embed dings, with heightened activity during this time period. However, speech embeddings from these multimodal models still show a similar alignment compared to their unimodal counter parts, suggesting that they do not gain meaning ful semantic benefits over text-based represen tations. These results highlight an asymmetry in cross-modal knowledge transfer, where the text modality benefits more from speech infor mation, but not vice versa. We make the code publicly available.

@inproceedings{bib_Ambi_2025, AUTHOR = {Baru, Lalith Bharadwaj and Dadi, Kamalaker and Chakraborty, Tapabrata and Surampudi, Bapi Raju }, TITLE = {Ambiguous Medical Image Segmentation Using Diffusion Schrödinger Bridge}, BOOKTITLE = {International Conference on Medical Image Computing and Computer Assisted Intervention}. YEAR = {2025}}

Accurate segmentation of medical images is challenging due to unclear lesion boundaries and mask variability. We introduce Segmentation Schödinger Bridge (SSB), the first application of Schödinger Bridge for ambiguous medical image segmentation, modelling joint image-mask dynamics to enhance performance. SSB preserves structural integrity, delineates unclear boundaries without additional guidance, and maintains diversity using a novel loss function. We further propose the Diversity Divergence Index () to quantify inter-rater variability, capturing both diversity and consensus. SSB achieves state-of-the-art performance on LIDC-IDRI, COCA, and RACER (in-house) datasets.

@inproceedings{bib_Corr_2025, AUTHOR = {Jindal, Akshett Rai and Reddy, Oota Subba and Mondal, Ishani and Pahwa, Khushbu and GNVV, Satya Sai Srinath Namburi and Shrivastava, Manish and Singh, Maneesh and Surampudi, Bapi Raju and Gupta, Manish }, TITLE = {Correlating instruction-tuning (in multimodal models) with vision-language processing (in the brain)}, BOOKTITLE = {International Conference on Learning Representations}. YEAR = {2025}}

Transformer-based language models, though not explicitly trained to mimic brain recordings, have demonstrated surprising alignment with brain activity. Progress in these models—through increased size, instruction-tuning, and multimodality—has led to better representational alignment with neural data. Recently, a new class of instruction-tuned multimodal LLMs (MLLMs) have emerged, showing remarkable zero-shot capabilities in open-ended multimodal vision tasks. However, it is unknown whether MLLMs, when prompted with natural instructions, lead to better brain alignment and effectively capture instruction-specific representations. To address this, we first investigate the brain alignment, i.e., measuring the degree of predictivity of neural visual activity using text output response embeddings from MLLMs as participants engage in watching natural scenes. Experiments with 10 different instructions (like image captioning, visual question answering, etc.) show that MLLMs exhibit significantly better brain alignment than vision-only models and perform comparably to non-instruction-tuned multimodal models like CLIP. We also find that while these MLLMs are effective at generating high-quality responses suitable to the task-specific instructions, not all instructions are relevant for brain alignment. Further, by varying instructions, we make the MLLMs encode instruction-specific visual concepts related to the input image. This analysis shows that MLLMs effectively capture count-related and recognition-related concepts, demonstrating strong alignment with brain activity. Notably, the majority of the explained variance of the brain encoding models is shared between MLLM embeddings of image captioning and other instructions. These results indicate that enhancing MLLMs' ability to capture more task-specific information could allow for better differentiation between various types of instructions, and hence improve their precision in predicting brain responses.

@inproceedings{bib_Feas_2024, AUTHOR = {Surampudi, Bapi Raju }, TITLE = {Feasibility and efficacy of virtual reality rehabilitation compared with conventional physiotherapy for upper extremity impairment due to ischaemic stroke: protocol for a randomised controlled trial }, BOOKTITLE = {BMJ}. YEAR = {2024}}

@inproceedings{bib_A_Co_2024, AUTHOR = {Vesapogu, Praveen Kumar and Surampudi, Bapi Raju }, TITLE = {A Comparative Study of Machine Learning Approaches for the Detection of SARS-CoV-2 and its Variants}, BOOKTITLE = {Procedia Computer Science}. YEAR = {2024}}

The coronavirus disease (COVID-19) caused over 170 million illnesses and over 3 million deaths worldwide. Researchers from all over the world have been using a variety of machine-learning techniques to classify the DNA sequences of the SARS-CoV-2 virus. To classify SARS-CoV-2 (Covid-19) whole genome sequences with other viruses including Dengue, Ebola, Influenza, and also Coronaviruses that may infect humans and are members of the family of Coronaviridae like SARS-CoV, MERS-CoV, we have proposed a multi-dimensional frequency encoding scheme. In the proposed method each genome is converted into four frequency-encoded sequences for each nucleotide. Then each encoded sequence is partitioned into quartiles and fed to the CNN. The findings of the binary classification of SARS-CoV-2 with other viruses and multi-class classification of six different viruses using five well-known machine learning algorithms and the proposed approach using CNN were reported in this study. Furthermore, we applied machine learning methods for the multi-class classification of eight different SARS-CoV-2 variants. The proposed approach performed well among all machine-learning techniques for classifying genome sequences of viruses and SARS-CoV-2 variants with accuracy of 99% and 98% respectively. In addition, we have also compared the performance of our proposed approach with the existing methods from the literature.

@inproceedings{bib_Mult_2024, AUTHOR = {Surampudi, Bapi Raju }, TITLE = {Multi-modal brain encoding models for multi-modal stimuli}, BOOKTITLE = {Technical Report}. YEAR = {2024}}

Despite participants engaging in single modality stimuli, such as watching images or silent videos, recent work has demonstrated that multi-modal Transformer models can predict visual brain activity impressively well, even with incongruent modality representations. This raises the question of how accurately these multi-modal models can predict brain activity when participants are engaged in multi-modal stimuli. As these models grow increasingly popular, their use in studying neural activity provides insights into how our brains respond to such multi-modal naturalistic stimuli, i.e., where it separates and integrates information from different sensory modalities. We investigate this question by using multiple unimodal and two types of multi-modal models—cross-modal and jointly pretrained—to determine which type of models is more relevant to fMRI brain activity when participants were engaged in watching movies (videos with audio). We observe that both types of multi-modal models show improved alignment in several language and visual regions. This study also helps in identifying which brain regions process unimodal versus multi-modal information. We further investigate the impact of removal of unimodal features from multi-modal representations and find that there is additional information beyond the unimodal embeddings that is processed in the visual and language regions. Based on this investigation, we find that while for cross-modal models, their brain alignment is partially attributed to the video modality; for jointly pretrained models, it is partially attributed to both the video and audio modalities.

@inproceedings{bib_Loca_2024, AUTHOR = {Shantilal, Jain Chirag and Sravanthi, Upadrasta Naga Sita and Sharma, Avinash and Surampudi, Bapi Raju }, TITLE = {Localizing the Sources of Diffusion Mediating Structure-Function Mapping Using Graph Diffusion Wavelets}, BOOKTITLE = {Network Neuroscience}. YEAR = {2024}}

The intricate link between brain textbf{Functional Connectivity (FC)} and textbf{Structural Connectivity (SC)} is explored through models performing diffusion on SC to derive FC, using varied methodologies from single to multiple textbf{Graph Diffusion Kernels}. However, existing studies have not correlated diffusion scales with specific brain Regions of Interest (RoIs), limiting the applicability of graph diffusion. We propose a novel approach using textbf{Graph Diffusion Wavelets} to learn the appropriate diffusion scale for each RoI to accurately estimate the SC-FC mapping. Using the open Human Connectome Project (HCP) dataset, we achieve an average textbf{Pearson's correlation} value of 0.833, surpassing the state-of-the-art methods for prediction of FC. It is important to note that the proposed architecture is entirely linear, computationally efficient, and notably demonstrates the textbf{power-law} distribution of diffusion scales. Our results show that the bilateral frontal pole, by virtue of it having large diffusion scale, forms a large community structure. The finding is in line with the current literature on the role of the frontal pole in resting-state networks. Overall, the results underscore the potential of graph diffusion wavelet framework for understanding how the brain structure leads to functional connectivity.

@inproceedings{bib_Hype_2024, AUTHOR = {Jain, Chirag and Arora, Mehul and Baru, Lalith Bharadwaj and Surampudi, Bapi Raju }, TITLE = {HyperGALE: ASD Classification via Hypergraph Gated Attention with Learnable Hyperedges}, BOOKTITLE = {International Joint Conference on Neural Networks}. YEAR = {2024}}

Autism Spectrum Disorder (ASD) is a neurodevelopmental condition characterized by varied social cognitive challanges and repetitive behaviour patterns. Identifying reliable brain imaging-based biomarkers for ASD has been a persistent challenge due to the spectrum's diverse symptomatology. Existing baselines in the field have made significant strides in this direction, yet there remains room for improvement in both performance and interpretability. We propose HyperGALE, which builds on the hypergraph by incorporating learned hyper graphs and gated attention mechanisms. This approach has led to substantial improvements in the model's ability to interpret complex brain graph data, offering deeper insights into ASD biomarker characterization. Evaluated on the exptensive ABIDE II dataset, HyperGALE not only improves interpretability but also demonstrates statistically significant enchancements in key performance metrics compared to both previous baselines and the foundation hypergraph model. The advancement HyperGALE brings to ASD research highlights the potential of sophisticated graph-based techniques in neurodevelopmental studies. The source code and implementation instructions are available at Github.

@inproceedings{bib_Atte_2024, AUTHOR = {Reddy, Oota Subba and Yadav, Archi and DASH, ARPITA and Surampudi, Bapi Raju and Sharma, Avinash }, TITLE = {Attention-based fusion of multiple graphheat networks for structural to functional brain mapping}, BOOKTITLE = {Scientific Reports}. YEAR = {2024}}

Over the last decade, there has been growing interest in learning the mapping from structural connectivity (SC) to functional connectivity (FC) of the brain. The spontaneous fuctuations of the brain activity during the restingstate as captured by functional MRI (rsfMRI) contain rich non stationary dynamics over a relatively fxed structural connectome. Among the modeling approaches, graph difusionbased methods with single and multiple difusion kernels approximating static or dynamic functional connectivity have shown promise in predicting the FC given the SC. However, these methods are computationally expensive, not scalable, and fail to capture the complex dynamics underlying the whole process. Recently, deep learning methods such as GraphHeat networks and graph difusion have been shown to handle complex relational structures while preserving global information. In this paper, we propose a novel attentionbased fusion of multiple GraphHeat networks (AGHN) for mapping SCFC. AGHN enables us to model multiple heat kernel difusion over the brain graph for approximating the complex Reaction Difusion phenomenon. We argue that the proposed deep learning method overcomes the scalability and computational inefciency issues but can still learn the SCFC mapping successfully. Training and testing were done using the rsfMRI data of 1058 participants from the human connectome project (HCP), and the results establish the viability of the proposed model. On HCP data, we achieve a high Pearson correlation of 0.788 (DesikanKilliany atlas with 87 regions) and 0.773 (AAL atlas with 86 regions). Furthermore, experiments demonstrate that AGHN outperforms the existing methods in learning the complex nature of the structurefunction relation of the human brain.

@inproceedings{bib_Tran_2024, AUTHOR = {Sharma, Shivam Kumar and Maiti, Suvadeep and Mythirayee, S. and Rajendran, Srijithesh and Surampudi, Bapi Raju }, TITLE = {Transparency in Sleep Staging: Deep Learning Method for EEG Sleep Stage Classification with Model Interpretability}, BOOKTITLE = {Technical Report}. YEAR = {2024}}

Automated Sleep stage classification using raw single channel EEG is a critical tool for sleep quality assessment and disorder diagnosis. However, modelling the complexity and variability inherent in this signal is a challenging task, limiting their practicality and effectiveness in clinical settings. To mitigate these challenges, this study presents an end-to-end deep learning (DL) model which integrates squeeze and excitation blocks within the residual network to extract features and stacked Bi-LSTM to understand complex temporal dependencies. A distinctive aspect of this study is the adaptation of GradCam for sleep staging, marking the first instance of an explainable DL model in this domain with alignment of its decision-making with sleep expert's insights. We evaluated our model on the publically available datasets (SleepEDF-20, SleepEDF-78, and SHHS), achieving Macro-F1 scores of 82.5, 78.9, and 81.9, respectively. Additionally, a novel training efficiency enhancement strategy was implemented by increasing stride size, leading to 8x faster training times with minimal impact on performance. Comparative analyses underscore our model outperforms all existing baselines, indicating its potential for clinical usage.

@inproceedings{bib_Enha_2024, AUTHOR = {Maiti, Suvadeep and Sharma, Shivam Kumar and Surampudi, Bapiraju }, TITLE = {Enhancing Healthcare with EOG: A Novel Approach to Sleep Stage Classification}, BOOKTITLE = {International Conference on Acoustics, Speech, and Signal Processing}. YEAR = {2024}}

We introduce an innovative approach to automated sleep stage classification using EOG signals, addressing the discomfort and impracticality associated with EEG data acquisition. In addition, it is important to note that this approach is untapped in the field, highlighting its potential for novel insights and contributions. Our proposed SE-Resnet-Transformer model provides an accurate classification of five distinct sleep stages from raw EOG signal. Extensive validation on publically available databases (SleepEDF-20, SleepEDF-78, and SHHS) reveals noteworthy performance, with macro-F1 scores of 74.72, 70.63, and 69.26, respectively. Our model excels in identifying REM sleep, a crucial aspect of sleep disorder investigations. We also provide insight into the internal mechanisms of our model using techniques such as 1D-GradCAM and t-SNE plots. Our method improves the accessibility of sleep stage classification while decreasing the need for EEG modalities. This development will have promising implications for healthcare and the incorporation of wearable technology into sleep studies, thereby advancing the field's potential for enhanced diagnostics and patient comfort.

@inproceedings{bib_BeSt_2023, AUTHOR = {Deb, Prantik and Baru, Lalith Bharadwaj and Dadi, Kamalaker and Surampudi, Bapi Raju }, TITLE = {BeSt-LeS: Benchmarking Stroke Lesion Segmentation using Deep Supervision}, BOOKTITLE = {International Conference on Medical Image Computing and Computer Assisted Intervention}. YEAR = {2023}}

Brain stroke has become a significant burden on global health and thus we need remedies and prevention strategies to overcome this challenge. For this, the immediate identification of stroke and risk stratification is the primary task for clinicians. To aid expert clinicians, automated segmentation models are crucial. In this work, we consider the publicly available dataset ATLAS v2.0 to benchmark various end-to-end supervised U-Net style models. Specifically, we have benchmarked models on both 2D and 3D brain images and evaluated them using standard metrics. We have achieved the highest Dice score of 0.583 on the 2D transformer-based model and 0.504 on the 3D residual U-Net respectively. We have conducted the Wilcoxon test for 3D models to correlate the relationship between predicted and actual stroke volume.

@inproceedings{bib_Enco_2023, AUTHOR = {, and Surampudi, Bapiraju }, TITLE = {Encoded Deep Vectors for Eukaryotic Exon Prediction}, BOOKTITLE = {Conference on Pattern Recognition and Machine Intelligence}. YEAR = {2023}}

@inproceedings{bib_Acti_2023, AUTHOR = {DONAPATI, ROHAN RAVI REDDY and Shukla, Anuj and Surampudi, Bapiraju }, TITLE = {Action-Outcome Delays Modulate the Temporal Expansion of Intended Outcomes}, BOOKTITLE = {NPG Nature Scientific Reports}. YEAR = {2023}}

The phenomenon of intentional binding pertains to the perceived connection between a voluntary action and its anticipated result. When an individual intends an outcome, it appears to subjectively extend in time due to a pre-activation of the intended result, particularly evident at shorter action-outcome delays. However, there is a concern that the operationalization of intention might have led to a mixed interpretation of the outcome expansion attributed to the pre-activation of intention, given the sensitivity of time perception and intentional binding to external cues that could accelerate the realization of expectations. To investigate the expansion dynamics of an intended outcome, we employed a modified version of the temporal bisection task in two experiments. Experiment 1 considered the action-outcome delay as a within-subject factor, while experiment 2 treated it as a between-subject factor. The results revealed that the temporal expansion of an intended outcome was only evident under the longer action-outcome delay condition. We attribute this observation to working memory demands and attentional allocation due to temporal relevancy, and not due to pre-activation. The discrepancy in effects across studies is explained by operationalizing different components of the intentional binding effect, guided by the cue integration theory. Moreover, we discussed speculative ideas regarding the involvement of specific intentions based on the proximal intent distal intent (PIDI) theory, and whether causality plays a role in temporal binding. Our study contributes to the understanding of how intention influences time perception and sheds light on how various methodological factors, cues, and delays can impact the dynamics of temporal expansion associated with an intended outcome.

@inproceedings{bib_Inte_2023, AUTHOR = {DONAPATI, ROHAN RAVI REDDY and Shukla, Anuj and Surampudi, Bapiraju }, TITLE = {Intended Outcomes Expand In Time: Evidence from the Temporal Reproduction Task}, BOOKTITLE = {Annual Meeting of the Cognitive Science Society}. YEAR = {2023}}

The intentional binding (IB) phenomenon reflects a perceived attraction between a voluntary action and an intendedconsequence. Recent research has shown that during IB, intended outcomes expand in time. Although, this effect was significant only for shorter action-outcome delays. However, literature on IB suggests it can also exist for long delays. To address this, we implemented a temporal reproduction task to observe the expansion of an intended outcome. Results revealed the expansion of an intended outcome in the shorter and longer action-outcome delays. These results were discussed under the PIDI (proximal intent distal intent) framework since, using this methodology, both types of intention can be operationalised. Proximal for shorter delays under the motor action, and distal for longer delays as the inference made towards the objective duration to be reproduced. Our implementation allows a robust way to observe the temporal dynamics of an intended outcome, irrespective of the action-outcome delay.

@inproceedings{bib_A_Ph_2023, AUTHOR = {Bandyopadhyay, Anirban and Ghosh, Sayan and Biswas, Dipayan and Surampudi, Bapiraju and Chakravarthy, V. Srinivasa }, TITLE = {A Phenomenological Deep Oscillatory Neural Network Model to Capture the Whole Brain Dynamics in Terms of BOLD Signal}, BOOKTITLE = {International Conference on Neural Information Processing}. YEAR = {2023}}

A large-scale model of brain dynamics, as it is manifested in functional neuroimaging data, is presented in this study. The model is built around a general trainable network of Hopf oscillators, the dynamics of which are described in the complex domain. It was shown earlier that when a pair of Hopf oscillators are coupled by power coupling with a complex coupling strength, it is possible to stabilize the normal phase difference at a value related to the angle of the complex coupling strength. In the present model, the magnitudes of the complex coupling weights are set using the Structural Connectivity information obtained from Diffusion Tensor Imaging (DTI). The complex-valued outputs of the oscillator network are transformed by a complex-valued feedforward network with a single hidden layer. The entire model is

@inproceedings{bib_Mode_2023, AUTHOR = {Venugopal, Gautham and Surampudi, Bapiraju }, TITLE = {Modelling Grid Navigation Using Reinforcement Learning Linear Ballistic Accumulators}, BOOKTITLE = {International Joint Conference on Neural Networks}. YEAR = {2023}}

Reinforcement Learning (RL) models constitute an important subset of models used in studying many facets of human learning including Motor Sequence Learning. However, conventional action selection in RL models such as softmax-based choice rules lack biological plausibility and do not offer mechanistic explanations. Furthermore, they also do not use response time data in model fitting, which can be indicative of the difference in value of alternate choices as perceived by subjects. Evidence Accumulation Models(EAM) such as Linear Ballistic Accumulators(LBA) provide a solution to many of the above problems. In this study, we use RL algorithms integrated with an LBA model to model human behaviour in the Grid-Sailing Task. The task involves navigating a grid to reach a goal position where the participant can choose from three possible actions. We fit RLLBA models using three different RL algorithms: a …

@inproceedings{bib_Dyna_2023, AUTHOR = {Prasad, Pindi Krishna Chandra and Dadi, Kamalakar and Surampudi, Bapiraju }, TITLE = {Dynamic functional connectivity analysis in individuals with Autism Spectrum Disorder}, BOOKTITLE = {International Joint Conference on Neural Networks}. YEAR = {2023}}

Autism spectrum disorder (ASD) is a neurodevelopmental disorder that predominantly occurs in children. Previous brain research in ASD has mainly studied biomarkers based on the functional connectivity characterized by the correlation of static temporal signals. However, brain connectivity is dynamic and varies extensively among brain states. The main aim of the paper is to understand the fundamental group differences between ASD patients and typically developing (TD) subjects using dynamic functional connectivity (dFNC) analysis. In this study, we investigated the dFNC between 53 independent components among 188 ASD and 195 TD subjects. We estimated dFNC using sliding window-based approaches and identified four distinct dynamic states through hard-clustering analysis. Hyper-connectivity within the cognitive control domain, between cognitive control and default mode network, has been …

@inproceedings{bib_Appl_2023, AUTHOR = {Sravanthi, Upadrasta Naga Sita and Dadi, Kamalakar and Surampudi, Bapiraju and PN.Sylaja, and Das, C.Kesava and Das, C.Kesava }, TITLE = {Application of Graph Theoretic measures for assessing efficacy of Stroke Rehabilitation}, BOOKTITLE = {International Joint Conference on Neural Networks}. YEAR = {2023}}

One of the challenges in stroke rehabilitation is to identify bio-markers that correlate with amelioratory changes in the recovery of brain function that can be verified by a clinician. For strokes related to upper extremity, clinicians use Fugl Meyer Assessment - Upper Extremity (FMA-UE) score for verification. We hypothesize that even before clinical measures of function recovery (FMA-UE) show facilitatory changes, structural and functional changes in the brain connectome indicate early changes in stroke patients. Toward establishing such early bio-markers of rehabilitation related changes in the brain plasticity, we propose to use graph theoretic measures on the structural (SC) and functional connectivity (FC) matrices. We used longitudinal multi-modal neuroimaging data acquired within 1 to 6 months of the onset of stroke and after 3 months of rehabilitation for 15 acute ischemic stroke subjects with deficit in motor

@inproceedings{bib_Emer_2023, AUTHOR = {Milagres, Fernandes Dolton and Kaushik, Pramod and Surampudi, Bapiraju }, TITLE = {Emergence of Direction Selectivity and Motion Strength in Dot Motion Task Through Deep Reinforcement Learning Networks}, BOOKTITLE = {International Joint Conference on Neural Networks}. YEAR = {2023}}

Deep Reinforcement learning is beginning to be useful for studying neural representations in the brain because of its ability to combine decision-making and representation. Here, we use it to study a dot motion perceptual decision-making task in a high-dimensional setting where the inputs are akin to those used in psychological experiments. This end-to-end model gives a unique insight into how these networks solve the task providing a background on how the brain could solve this task. We find that the network can show properties similar to the middle temporal visual area (MT) in the brain, which code for direction and motion strength. We find the emergence of direction selectivity purely through reward-based training and graded firing coding motion strength and make a testable prediction that the MT population would also have coherence-selective neurons.

@inproceedings{bib_How__2023, AUTHOR = {Oota, Subba Reddy and Marreddy, Mounika and Gupta, Manish and Surampudi, Bapiraju }, TITLE = {How does the brain process syntactic structure while listening?}, BOOKTITLE = {Findings of the Association for Computational Linguistics}. YEAR = {2023}}

Syntactic parsing is the task of assigning a syntactic structure to a sentence. There are two popular syntactic parsing methods: constituency and dependency parsing. Recent works have used syntactic embeddings based on constituency trees, incremental top-down parsing, and other word syntactic features for brain activity prediction given the text stimuli to study how the syntax structure is represented in the brain’s language network. However, the effectiveness of dependency parse trees or the relative predictive power of the various syntax parsers across brain areas, especially for the listening task, is yet unexplored. In this study, we investigate the predictive power of the brain encoding models in three settings: (i) individual performance of the constituency and dependency syntactic parsing based embedding methods, (ii) efficacy of these syntactic parsing based embedding methods when controlling for basic syntactic signals, (iii) relative effectiveness of each of the syntactic embedding methods when controlling for the other. Further, we explore the relative importance of syntactic information (from these syntactic embedding methods) versus semantic information using BERT embeddings. We find that constituency parsers help explain activations in the temporal lobe and middle-frontal gyrus, while dependency parsers better encode syntactic structure in the angular gyrus and posterior cingulate cortex. Although semantic signals from BERT are more effective compared to any of the syntactic features or embedding methods, syntactic embedding methods explain additional v

@inproceedings{bib_Deep_2023, AUTHOR = {OOta, Subba Rao and Gupta, manish and Surampudi, Bapiraju and Jobard, Gael and Alexandre, Frederic and Hinaut, Xavier }, TITLE = {Deep Neural Networks and Brain Alignment: Brain Encoding and Decoding (Survey)}, BOOKTITLE = {Technical Report}. YEAR = {2023}}

How does the brain represent different modes of information? Can we design a system that automatically understands what the user is thinking? Such questions can be answered by studying brain recordings like functional magnetic resonance imaging (fMRI). As a first step, the neuroscience community has contributed several large cognitive neuroscience datasets related to passive reading/listening/viewing of concept words, narratives, pictures and movies. Encoding and decoding models using these datasets have also been proposed in the past two decades. These models serve as additional tools for basic research in cognitive science and neuroscience. Encoding models aim at generating fMRI brain representations given a stimulus automatically. They have several practical applications in evaluating and diagnosing neurological conditions and thus also help design therapies for brain damage. Decoding models solve the inverse problem of reconstructing the stimuli given the fMRI. They are useful for designing brain-machine or brain-computer interfaces. Inspired by the effectiveness of deep learning models for natural language processing, computer vision, and speech, recently several neural encoding and decoding models have been proposed. In this survey, we will first discuss popular representations of language, vision and speech stimuli, and present a summary of neuroscience datasets. Further, we will review popular deep learning based encoding and decoding architectures and note their benefits and limitations. Finally, we will conclude with a brief summary and discussion about future trends

@inproceedings{bib_Spee_2023, AUTHOR = {OOta, Subbareddy and Agarwal, Veeral and Marreddy, Mounika and Gupta, Manish and Surampudi, Bapiraju }, TITLE = {Speech Taskonomy: Which Speech Tasks are the most Predictive of fMRI Brain Activity?}, BOOKTITLE = {Annual Conference of the International Speech Communication Association}. YEAR = {2023}}

Self-supervised speech based models have been found to be successful in predicting brain recordings of subjects experiencing naturalistic story listening. Inspired by the recent progress on deep learning models for various speech-processing tasks, existing literature has leveraged pretrained speech Transformer models for brain encoding. However, there is no work on exploring the efficacy of task-specific finetuned Transformer representations for this task. Hence, in this paper, we explore transfer learning from representations finetuned for eight different tasks from Speech processing Universal PERformance Benchmark (SUPERB) for predicting brain responses. Encoding models based on task features are used to predict activity in different regions across the whole brain, and also in language and auditory brain regions. Our experiments on finetuning the Wav2Vec2.0 model for these eight tasks show that the model finetuned on automatic speech recognition (ASR) yields the best encoding performance for the whole brain, language and auditory regions.

@inproceedings{bib_Leak_2023, AUTHOR = {Anamalamud, Komala and Surampudi, Bapiraju and Chakraborty, Goutam and Vakkala, Nirupa }, TITLE = {Leaky-Integrate-and-Fire Neuron as Pacemaker for Interval Timing}, BOOKTITLE = {IEEE Delhi Section Conference}. YEAR = {2023}}

Perception of interval timing influences the behaviour of the organisms. Computational models of interval timing are categorized into Pacemaker Accumulator models, Memory-based models, Oscillator models and Random Process models, Ramping Activity models and Population Clock models. Random process models or drift diffusion models are biologically plausible models and are based on the activity of spiking neurons. In this paper, we proposed a computational model of interval timing based on spiking neurons. The results are validated against the Scalar property of interval timing

@inproceedings{bib_Subj_2023, AUTHOR = {Surampudi, Bapiraju }, TITLE = {Subjective Time Estimation to Measure the Cognitive Load of Interactive Mobile User Interfaces}, BOOKTITLE = {International Conference onIntelligent Systems for Communication, IoT and Security}. YEAR = {2023}}

Measuring subjective experience of time helps the designer of the product to understand the usability of the product and the user experience. Time is not given it’s due importance in HCI and particularly in the usability design of the products. Though it is evident that cognitive load impacts time/duration estimation, thereby the user creates a positive or negative valence towards the product, there is no much research done to measure the usability of the products with cognitive load and duration judgement as dependent variables. This paper focuses on empirical investigation of prospective duration judgement as a measure to assess the cognitive load experienced by the users while using the product, one-on-one mentoring mobile app. We consider Jakob Nielsen’s heuristics and meCUE questionnaire as the anchors for our empirical investigation. Index Terms—Human-Computer Interaction, Duration Judge- ment, Usability, User Experience,

@inproceedings{bib_Big__2023, AUTHOR = {Dadi, Kamalaker and Surampudi, Bapiraju }, TITLE = {Big Data in Cognitive Neuroscience: Opportunities and Challenges}, BOOKTITLE = {International Conference on Big Data Analytics}. YEAR = {2023}}

Cognitive brain mapping is enjoying its growth with the availability of large open data sharing efforts as well as the application of modern machine learning and deep learning methods. In this article, we review some of the current practices in cognitive neuroscience predominantly focusing on functional imaging and highlight the tremendous opportunities fostered by the unprecedented scale of datasets in cognitive neuroscience. We also point out challenges and limitations to keep in mind while working with these datasets.

@inproceedings{bib_Mome_2022, AUTHOR = {Milagres, Fernandes Dolton and KAUSHIK, PRAMOD SIVARAM and Shukla, Harsh and Surampudi, Bapiraju }, TITLE = {Momentum Boosted Episodic Memory for Improving Learning in Long-Tailed RL Environments}, BOOKTITLE = {Neural Information Processing Systems Workshops}. YEAR = {2022}}

Conventional Reinforcement Learning (RL) algorithms assume the distribution of the data to be uniform or mostly uniform. However, this is not the case with most real-world applications like autonomous driving or in nature, where animals roam. Some objects are encountered frequently, and most of the remaining experiences occur rarely; the resulting distribution is called Zipfian. Taking inspiration from the theory of complementary learning systems, an architecture for learning from Zipfian distributions is proposed where long tail states are discovered in an unsupervised manner and states along with their recurrent activation are kept longer in episodic memory. The recurrent activations are then reinstated from episodic memory using a similarity search, giving weighted importance. The proposed architecture yields improved performance in a Zipfian task over conventional architectures. Our method outperforms IMPALA by a significant margin of 20.3% when maps/objects occur with a uniform distribution and by 50.2% on the rarest 20% of the distribution.

@inproceedings{bib_Priv_2022, AUTHOR = {Gadepaka, Latha and Surampudi, Bapiraju }, TITLE = {Privacy Preserving of Two Collaborating Parties Using Fuzzy C-Means Clustering}, BOOKTITLE = {International Conference on Advanced Network Technologies and Intelligent Computing}. YEAR = {2022}}

In this smart and rapidly growing computing world, most of the organizations and companies needs to share necessary information (data) to third parties for their analysis which plays major role in decision making. Any data generally consists of sensitive (personal) information about people and organizations and when the sensitive information to be revealed to third parties, there is always chance of privacy loss. In this regard any organization or an individual requires the application of privacy preserving techniques to ensure data privacy, that means the data to be exchanged between two or more sites, without data loss or privacy violation. When data to

@inproceedings{bib_Deep_2022, AUTHOR = {Oota, Subba Reddy and Arora, Jashn and Gupta, Manish and Surampudi, Bapiraju and Toneva, Mariya }, TITLE = {Deep Learning for Brain Encoding and Decoding}, BOOKTITLE = {Annual Meeting of the Cognitive Science Society}. YEAR = {2022}}

How does the brain represent different modes of information? Can we design a system that can automatically understand what the user is thinking? We can make progress towards answering such questions by studying brain recordings from devices such as functional magnetic resonance imaging (fMRI). The brain encoding problem aims to automatically generate fMRI brain representations given a stimulus. The brain decoding problem is the inverse problem of reconstructing the stimuli given the fMRI brain representation. Both the brain encoding and decoding problems have been studied in detail in the past two decades and the foremost attraction of studying these solutions is that they serve as additional tools for basic research in cognitive science and cognitive neuroscience. Recently, inspired by the effectiveness of deep learning models for natural language processing and computer vision, such models have been applied for neuroscience as well. In this tutorial, we plan to discuss different kinds of stimulus representations, and popular encoding and decoding architectures in detail. The tutorial will provide a working knowledge of the state of the art methods for encoding and decoding, a thorough understanding of the literature, and a better understanding of the benefits and limitations of encoding/decoding with deep learning. Encoding models that accurately predict brain activity have several practical applications in evaluation and diagnosis of neurological conditions and thus also help designing therapies for brain damage. Invertible encoding models enable principled formulation of brain decoding models which in turn are useful for designing brain-machine or brain-computer interfaces. Recent advances in the use of pretrained deep network models enable us to use them as priors for brain decoding tasks. Deep learning models are useful for improving accuracy but also offer the flexibility of decoding across a gamut of tasks and domains.

@inproceedings{bib_Priv_2022, AUTHOR = {Gadepaka, Latha and Surampudi, Bapiraju }, TITLE = {Privacy Preserving of Two Collaborating Parties Using Fuzzy C-Means Clustering}, BOOKTITLE = {International Conference on Advanced Network Technologies and Intelligent Computing}. YEAR = {2022}}

In this smart and rapidly growing computing world, most of the organizations and companies needs to share necessary information (data) to third parties for their analysis which plays major role in decision making. Any data generally consists of sensitive (personal) information about people and organizations and when the sensitive information to be revealed to third parties, there is always chance of privacy loss. In this regard any organization or an individual requires the application of privacy preserving techniques to ensure data privacy, that means the data to be exchanged between two or more sites, without data loss or privacy violation. When data to be shared between multiple organizations or sites (parties) the privacy issues will be arises more. In this line we choose collaborative fuzzy clustering that gives better solutions to preserve privacy of data while sharing information between multiple parties to achieve …

@inproceedings{bib_Brai_2022, AUTHOR = {Arora, Jashn and Reddy, Oota Subba and Gupta, Manish and Surampudi, Bapiraju }, TITLE = {Brain Decoding for Abstract versus Concrete Concepts}, BOOKTITLE = {Annual Conference of Cognitive Science}. YEAR = {2022}}

Neuroimaging techniques such as fMRI record brain activation while participants experience a stimulus. The concreteness of concepts defines how well our brain is able to imagine them. We hypothesise that brain activation would be distinctly different when participants view stimuli corresponding to concrete words versus abstract words. Specifically, we expect primary sensory areas to engage more in the former. To study this, we used a multi-view dataset with each concept being displayed as a picture, as a word in a sentence and also as a wordcloud [1]. We investigate how well a machine learning-based decoder trained on concrete concepts can decode both concrete and abstract concepts and vice versa for the model trained on abstract concepts. We find that a model trained on stimuli related to concrete concepts yields better accuracy than the model trained on abstract concepts. We also explore the contribution of voxels from different brain regions in this decoding process. Analysis of the contribution of different brain networks reveals exciting cognitive insights.

@inproceedings{bib_LRH-_2022, AUTHOR = {Chauhan, Ekansh and Guptha, Swathi and Reddy, Likith and Surampudi, Bapiraju }, TITLE = {LRH-Net: A Multi-level Knowledge Distillation Approach for Low-Resource Heart Network}, BOOKTITLE = {International Workshop on Distributed, Collaborative, and Federated Learning}. YEAR = {2022}}

An electrocardiogram (ECG) monitors the electrical activity generated by the heart and is used to detect fatal cardiovascular diseases (CVDs). Conventionally, to capture the precise electrical activity, clinical experts use multiple-lead ECGs (typically 12 leads). Recently, large-scale deep learning models have been used to detect these diseases, however, they require large memory and long inference time. We propose a low-parameter model, Low Resource Heart-Network (LRH-Net), that detects ECG anomalies in a resource-constrained environment. On top, multi-level knowledge distillation (MLKD) is employed to improve model generalization. MLKD distils the dark-knowledge from higher parameter models (teachers) trained on different lead configurations to LRH-Net. The LRH-Net has 106× fewer parameters and 76% faster inference than the teacher model for detecting CVDs. Using MLKD, the performance of LRH-Net on reduced lead data was scaled up to 3.25%, making it suitable for edge devices.

@inproceedings{bib_Rela_2022, AUTHOR = {SHUKLA, ANUJ KUMAR and Surampudi, Bapiraju }, TITLE = {Relative Numerical Context Affects Temporal Processing}, BOOKTITLE = {Annual Meeting of the Cognitive Science Society}. YEAR = {2022}}

Several studies have reported that numerical magnitudes biases temporal judgments, i.e., large numerical magnitude, were perceived to last longer than small numerical magnitude. However, these predictions have been predominantly verified only when the large and small numerical magnitudes were presented in an intermixed fashion where numerical magnitudes varied randomly from trial to trial. We conducted two experiments (Blockedmagnitude and Mixed-Magnitude) using a temporal bisection paradigm to investigate whether numerical context affects temporal processing in a sub-second timescale. The numbers were presented with varying durations. Participants were asked to judge whether the presented durations were shorter or longer. The results suggest that the temporal judgments were affected when small and large numbers were randomly presented in an intermixed manner. However, such effects disappeared when the number magnitudes were presented separately. These results indicate the modulation of attention in number-time interaction, and such crosstalk may not require a generalized magnitude system.

@inproceedings{bib_Numb_2022, AUTHOR = {SHUKLA, ANUJ KUMAR and Surampudi, Bapiraju }, TITLE = {Number-time interaction: Search for a common magnitude system in a cross-modal setting}, BOOKTITLE = {Frontiers in Behavioral Neuroscience}. YEAR = {2022}}

A theory of magnitude (ATOM) suggests that a generalized magnitude system in the brain processes magnitudes such as space, time, and numbers. Numerous behavioral and neurocognitive studies have provided support to ATOM theory. However, the evidence for common magnitude processing primarily comes from the studies in which numerical and temporal information are presented visually. Our current understanding of such cross-dimensional magnitude interactions is limited to visual modality only. However, it is still unclear whether the ATOM-framework accounts for the integration of cross-modal magnitude information. To examine the crossmodal influence of numerical magnitude on temporal processing of the tone, we conducted three experiments using a temporal bisection task. We presented the numerical magnitude information in the visual domain and the temporal information in the auditory either simultaneously with duration judgment task (Experiment-1), before duration judgment task (Experiment2), and before duration judgment task but with numerical magnitude also being task-relevant (Experiment-3). The results suggest that the numerical information presented in the visual domain affects temporal processing of the tone only when the numerical magnitudes were task-relevant and available while making a temporal judgment (Experiments-1 and 3). However, numerical information did not interfere with temporal information when presented temporally separated from the duration information (Experiments2). The findings indicate that the influence of visual numbers on temporal processing in cross-modal settings may not arise from the common magnitude system but instead from general cognitive mechanisms like attention and memory.

@inproceedings{bib_Priv_2022, AUTHOR = {Gadepaka, Latha and Surampudi, Bapiraju }, TITLE = {Privacy Preserving of Two Local Data Sites Using Global Random Forest Classification}, BOOKTITLE = {International Conference on Cyber Warfare, Security and Space Research}. YEAR = {2022}}

n the rapidly growing computing world, the data is being generated every minute, and it has been a big challenge to manage and maintain privacy of the data when it is required to share among different locations. In this regard the privacy preserving data mining is playing a major role to design and develop the methods and models to preserve privacy of data in data owners perspectives, when ever the necessity of exchanging the information between multiple parties to be done. There has been lot of research going on in the field of computational intelligence, machine learning and soft computing domains and some good number of algorithms and secured models are being developed by the researchers. We have worked on the random forest classification as the best example model to preserve privacy of two parties when a combined classification to be done in order to build a privacy preserving global random …

@inproceedings{bib_Priv_2022, AUTHOR = {Gadepaka, Latha and Surampudi, Bapiraju }, TITLE = {Privacy Preserving and Secured Clustering of Distributed Data Using Self Organizing Map}, BOOKTITLE = {International Conference on Cyber Warfare, Security and Space Research}. YEAR = {2022}}

Preserving privacy of data in data mining domain has been a challenging objective to achieve. There has been lot of progress in applying secure algorithms in order to achieve privacy preserving of sensitive information. In distributed data mining perspective, there is necessity to share the information among the different parties aiming for combined outcomes. Privacy preserving methods and algorithms helps in preserving privacy of data when it is compulsory to share the information between the collaborated parties to perform various data mining operations like Classification and Clustering. It is always important to address the privacy issues of an individual when the data is shared among different parties to achieve a common goal. It is observed that there has been great development in addressing the privacy issues and designing privacy preserving algorithms applied to achieve a secured model. Adopting neural …

@inproceedings{bib_A_VT_2022, AUTHOR = {KAUSHIK, PRAMOD SIVARAM and Naudé, Jérémie and Surampudi, Bapiraju and Alexandre, Frédéric }, TITLE = {A VTA GABAergic computational model of dissociated reward prediction error computation in classical conditioning}, BOOKTITLE = {Neurobiology of Learning and Memory}. YEAR = {2022}}

Classical Conditioning is a fundamental learning mechanism where the Ventral 14 Striatum is generally thought to be the source of inhibition to Ventral Tegmental Area (VTA) 15 Dopamine neurons when a reward is expected. However, recent evidences point to a new 16 candidate in VTA GABA encoding expectation for computing the reward prediction error in the VTA. 17 In this system-level computational model, the VTA GABA signal is hypothesised to be a combination 18 of magnitude and timing computed in the Peduncolopontine and Ventral Striatum respectively. 19 This dissociation enables the model to explain recent results wherein Ventral Striatum lesions 20 affected the temporal expectation of the reward but the magnitude of the reward was intact. This 21 model also exhibits other features in classical conditioning namely, progressively decreasing firing 22 for early rewards closer to the actual reward, twin peaks of VTA dopamine during training and 23 cancellation of US dopamine after training

@inproceedings{bib_Mult_2022, AUTHOR = {Reddy, Oota Subba and Arora, Jashn and Gupta, Manish and Surampudi, Bapiraju }, TITLE = {Multi-view and Cross-view Brain Decoding}, BOOKTITLE = {International Conference on Computational Linguistics}. YEAR = {2022}}

Can we build multi-view decoders that can decode concepts from brain recordings corresponding to any view (picture, sentence, word cloud) of stimuli? Can we build a system that can use brain recordings to automatically describe what a subject is watching using keywords or sentences? How about a system that can automatically extract important keywords from sentences that a subject is reading? Previous brain decoding efforts have focused only on single view analysis and hence cannot help us build such systems. As a first step toward building such systems, inspired by Natural Language Processing literature on multilingual and cross-lingual modeling, we propose two novel brain decoding setups: (1) multiview decoding (MVD) and (2) cross-view decoding (CVD). In MVD, the goal is to build an MV decoder that can take brain recordings for any view as input and predict the concept. In CVD, the goal is to train a model which takes brain recordings for one view as input and decodes a semantic vector representation of another view. Specifically, we study practically useful CVD tasks like image captioning, image tagging, keyword extraction, and sentence formation.

@inproceedings{bib_Bott_2022, AUTHOR = {Surampudi, Bapiraju and Janapati, Ravichander and Sengupta, Rakesh }, TITLE = {Bottom-up approach for enumeration of images using an on-center off-surround Recurrent Neural Network model}, BOOKTITLE = {Journal of Vision}. YEAR = {2022}}

Many studies on animals, children and adults advocate for an inherent mechanism for number perception in humans and animal brains. There are some computational models for visual number perception. The on-centre off-surround Recurrent Neural Network (RNN) From Sengupta et al.(2014) accounts for some major behavioural findings regarding number perception. The RNN entails an abstract form of input and output. In our current work, we have devised a biologically interpretable method to use real images as the network's input and retrieve numerical estimation from the network's output (mean activation), resulting in an end-to-end model for enumeration of images. The model uses information from Saliency maps of images as input current for the neurons for a small duration (presentation time). The activation of the neurons changes based on the internal dynamics of the network, characterised by self …

@inproceedings{bib_Inte_2022, AUTHOR = {Jain, Srishti and Surampudi, Bapiraju and Reeder, Reshanne and Sengupta, Rakesh }, TITLE = {Interference in Recall for the Features of a Single Object Upon Repeated Probes}, BOOKTITLE = {Journal of Vision}. YEAR = {2022}}

The interference model supports a limited capacity of working memory (WM): if multiple to-be-remembered objects are presented in succession, features of later objects interfere with the recall of features of earlier objects. Feature recall of earlier objects can be strengthened with rehearsal, but this is possible for a limited amount of information. We investigated whether rehearsal can reduce interference in the recall of features of a single object. In the current study (N= 24), a circle was presented with randomly varying color, size, and location over 400 trials. Within a single trial, the stimulus appeared followed by two sequential response screens. We probed the recall of color and location of the circle, and considered two conditions (repeat, non-repeat) to investigate recall accuracy. In the repeat condition, participants recalled one of the features (either color or location) twice in a row. In the non-repeat condition, each …

@inproceedings{bib_Ayu-_2022, AUTHOR = {Borkar, Kushal and Chaturvedi, Anusha and Krishnanunni, Vinod Palakkad and Surampudi, Bapiraju }, TITLE = {Ayu-Characterization of healthy aging from neuroimaging data with deep learning and rsfMRI}, BOOKTITLE = {Frontiers in Computational Neuroscience}. YEAR = {2022}}

Estimating brain age and establishing functional biomarkers that are prescient of cognitive declines resulting from aging and dierent neurological diseases are still open research problems. Functional measures such as functional connectivity are gaining interest as potentially more subtle markers of neurodegeneration. However, brain functions are also aected by “normal” brain aging. More information is needed on how functional connectivity relates to aging, particularly in the absence of neurodegenerative disorders. Resting-state fMRI enables us to investigate functional brain networks and can potentially help us understand the processes of development as well as aging in terms of how functional connectivity (FC) matures during the early years and declines during the late years. We propose models for estimation of the chronological age of a healthy person from the resting state brain activation (rsfMRI). In this work, we utilized a dataset (N = 638, age-range 20–88) comprising rsfMRI images from the Cambridge Centre for Aging and Neuroscience (Cam-CAN) repository of a healthy population. We propose an age prediction pipeline Ayu which consists of data preprocessing, feature selection, and an attention-based model for deep learning architecture for brain age assessment. We extracted features from the static functional connectivity (sFC) to predict the subject’s age and classified them into dierent age groups (young, middle, middle, and old ages). To the best of our knowledge, a classification accuracy of 72.619 % and a mean absolute error of 6.797, and an r 2 of 0.754 reported by our Ayu pipeline establish competitive benchmark results as compared to the state-of-the-art-approach. Furthermore, it is vital to identify how dierent functional regions of the brain are correlated. We also analyzed how functional regions contribute dierently across ages by applying attention-based networks and integrated gradients. We obtained well-known resting-state networks using the attention model, which maps to within the default mode network, visual network, ventral attention network, limbic network, frontoparietal network, and somatosensory network connected to aging. Our analysis of fMRI data in healthy elderly Age groups revealed that dynamic FC tends to slow down and becomes less complex and more random with increasing

@inproceedings{bib_Char_2022, AUTHOR = {DASH, ARPITA and Surampudi, Bapiraju and Roy, Dipanjan and Krishnanunni, Vinod Palakkad }, TITLE = {Characterizing the Dynamic Reorganization in Healthy Ageing and Classification of Brain Age}, BOOKTITLE = {International Joint Conference on Neural Networks}. YEAR = {2022}}

During healthy ageing, the brain networks undergo various topological and functional alterations. Previous studies have shown that the dedifferentiation of the functional modules could be one of the hallmarks of large-scale brain networks and alterations through the lifespan. This modular organiza- tion and alterations may be critically linked to a variety of neurodegenerative disorders and cognitive deficits encountered during ageing. In spite of accumulating evidence based on tracking static functional connectivity (FC) and modularity in characterizing dedifferentiation associated with ageing, there is a gap in understanding the brain dynamics of modular segregation and integration through the lifespan. Using the Cam-CAN dataset (young: 18-44, mean 32 years, old: 65-88, mean 75 years), we characterize the modular reorganization using dynamic measures like flexibility, to find characteristic nodes that make up the stable core and flexible periphery in the young and old age groups. In this study, we hypothesize that the nodes that exhibit higher flexibility in the older age groups will be negatively correlated with modularity since these nodes ‘compensate’ for the functional integration while ensuring that the segregation is efficient. Our results demonstrate that the regions from the Default Mode network (DMN) show a negative correlation with modularity in the old age groups. Further, nodes from Limbic, SensoriMotor (SMN) and Salience networks show a positive correlation with modularity. These networks that are responsible for higher-order cognitive functions, e.g., decision making, attentional control, cognitive flexibility are found to make up a stable core as evidenced by their low flexibility scores. We also trained various classifiers using node flexibility scores as features for the binary (young vs old) classification task. Support Vector Machine (SVM) with Gaussian kernel trained on a reduced-dimensional feature set gave the best classification results. The features (nodes) that are found to be important for classification concur with those identified through the data-driven network measures based analysis. In summary, we anticipate that these findings can help identify the regions that are responsible for the reorganization

@inproceedings{bib_Deep_2022, AUTHOR = {Prasad, Pindi Krishna Chandra and Khare, Yash and kamalaker.dadI, and Krishnanunni, Vinod Palakkad and Surampudi, Bapiraju }, TITLE = {Deep Learning Approach for Classification and Interpretation of Autism Spectrum Disorder}, BOOKTITLE = {International Joint Conference on Neural Networks}. YEAR = {2022}}

Autism spectrum disorder (ASD) is a neurodevelopmental disorder predominantly found in children. The current behavior-based diagnosis of ASD is arduous and requires expertise. Therefore, it is appealing to develop an accurate computer-aided tool for diagnosing ASD. Although resting-state functional magnetic resonance imaging (rsfMRI) has proven to be successful in capturing the neural organization of the brain, automated detection of ASD using rsfMRI scans is a challenging task due to heterogeneity in the dataset and limited sample size. This paper proposes a Multilayer Perceptron (MLP) based classification model with auto encoder pretraining for classifying ASD from Typically Developing (TD) using rsfMRI scans obtained from the ABIDE-1 dataset. Our model achieves new state-of-the-art performance on the ABIDE-1 dataset with a 10-fold cross-validation accuracy of 74.82%. Further, we use the Integrated Gradients (IG) and DeepLIFT techniques to identify the correlations between brain regions that contribute most to the classification task. Our analysis identifies the following regions, Left Lingual Gyrus, Right Insula Lobe, Right Cuneus, Right Middle Frontal Gyrus, Left Superior Temporal Gyrus to be associated with ASD. Interestingly, these regions in the brain are primarily responsible for social cognition, language, attention, decision making and visual processing, which are known to be altered in ASD.

@inproceedings{bib_Mult_2022, AUTHOR = {Latheef, Ammar Ahmed Pallikonda and Yadav, Archi and Sharma, Avinash and Surampudi, Bapiraju }, TITLE = {Multiple Kernel Learning for Modeling Resting State EEG Connectomes using Structural Connectivity of the Brain}, BOOKTITLE = {International Joint Conference on Neural Networks}. YEAR = {2022}}

An active area of research in cognitive science is characterizing the relationship between brain structure and the observed functional activations. Recent graph diffusion models have had great success in mapping whole-brain, resting-state dy- namics measured using functional Magnetic Resonance Imaging (fMRI) to the brain structure derived using diffusion and T1 brain imaging. Here we test the application of one such graph diffusion method called the Multiple Kernel Learning (MKL) model. MKL model, formulated as a reaction-diffusion system using Wilson-Cowan equations, combines multiple diffusion ker- nels at different scales to predict functional connectome (FC) arising from a fixed structural connectome (SC). Our simulation results demonstrate that the MKL model successfully mapped the relationship between SC and FC from five different Electroen- cephalogram (EEG) bands (delta, theta, alpha, beta, and gamma). We used simultaneously acquired EEG-fMRI and NODDI dataset of 17 participants. The correlation between predicted FC and ground truth FC was higher for EEG bands than for fMRI data. The prediction accuracy peaked for the alpha band, and the highest frequency band, gamma had the lowest prediction accuracy. To the best of our knowledge, this is the first such end- to-end application of multiple kernel graph diffusion framework for modeling EEG data. One of the important features of MKL model is its ability to incorporate structural connectivity features into the generative model that predicts the EEG functional connectivity.

@inproceedings{bib_Mult_2022, AUTHOR = {Reddy, Oota Subba and Yadav, Archi and DASH, ARPITA and Surampudi, Bapiraju and Sharma, Avinash }, TITLE = {Multiple GraphHeat Networks for Structural to Functional Brain Mapping}, BOOKTITLE = {International Joint Conference on Neural Networks}. YEAR = {2022}}

Over the last decade, there has been growing interest in learning the mapping from structural connectivity (SC) to functional connectivity (FC) of the brain. The spontaneous brain activity fluctuations during the resting-state as captured by func- tional MRI (rsfMRI) contain rich non-stationary dynamics over a relatively fixed structural connectome. Among the modeling approaches, graph diffusion-based methods with single and mul- tiple diffusion kernels approximating static or dynamic functional connectivity have shown promise in predicting the FC given the SC. However, these methods are computationally expensive, not scalable, and fail to capture the complex dynamics underlying the whole process. Recently, deep learning methods such as GraphHeat networks along with graph diffusion have been shown to handle complex relational structures while preserving global information. In this paper, we propose multiple GraphHeat networks (M-GHN), a novel approach for mapping SC-FC. M- GHN enables us to model multiple heat kernel diffusion over the brain graph for approximating the complex Reaction Diffusion phenomenon. We argue that the proposed deep learning method overcomes the scalability and computational inefficiency issues but can still learn the SC-FC mapping successfully. Training and testing were done using the rsfMRI data of 100 participants from the human connectome project (HCP), and the results establish the viability of the proposed model. On the HCP dataset of 100 participants, the M-GHN achieves a high Pearson correlation

@inproceedings{bib_Neur_2022, AUTHOR = {Reddy, Oota Subba and Arora, Jashn and Agarwal, Veeral and Marreddy, Mounika and Gupta, Manish and Surampudi, Bapiraju }, TITLE = {Neural Language Taskonomy: Which NLP Tasks are the most Predictive of fMRI Brain Activity?}, BOOKTITLE = {North American Chapter of the Association for Computational Linguistics: Human Language Technologies}. YEAR = {2022}}

Several popular Transformer based language models have been found to be successful for text-driven brain encoding. However, existing literature leverages only pretrained text Transformer models and has not explored the efficacy of task-specific learned Transformer representations. In this work, we explore transfer learning from representations learned for ten popular natural language processing tasks (two syntactic and eight semantic) for predicting brain responses from two diverse datasets: Pereira (subjects reading sentences from paragraphs) and Narratives (subjects listening to the spoken stories). Encoding models based on task features are used to predict activity in different regions across the whole brain. Features from coreference resolution, NER, and shallow syntax parsing explain greater variance for the reading activity. On the other hand, for the listening activity, tasks such as paraphrase generation, summarization, and natural language inference show better encoding performance. Experiments across all 10 task representations provide the following cognitive insights: (i) language left hemisphere has higher predictive brain activity versus language right hemisphere, (ii) posterior medial cortex, temporoparieto-occipital junction, dorsal frontal lobe have higher correlation versus early auditory and auditory association cortex, (iii) syntactic and semantic tasks display a good predictive performance across brain region

@inproceedings{bib_Cros_2022, AUTHOR = {Reddy, Oota Subba and Arora, Jashn and Gupta, Manish and Surampudi, Bapiraju }, TITLE = {Cross-view Brain Decoding}, BOOKTITLE = {Technical Report}. YEAR = {2022}}

How the brain captures the meaning of linguistic stimuli across multiple views is still a critical open question in neuroscience. Consider three different views of the concept apartment: (1) picture (WP) presented with the target word label, (2) sentence (S) using the target word, and (3) word cloud (WC) containing the target word along with other semantically related words. Unlike previous efforts, which focus only on single view analysis, in this paper, we study the effectiveness of brain decoding in a zero-shot cross-view learning setup. Further, we propose brain decoding in the novel context of cross-view-translation tasks like image captioning (IC), image tagging (IT), keyword extraction (KE), and sentence formation (SF). Using extensive experiments, we demonstrate that cross-view zero-shot brain decoding is practical leading to ∼0.68 average pairwise accuracy across view pairs. Also, the decoded representations are sufficiently detailed to enable high accuracy for cross-view-translation tasks with following pairwise accuracy: IC (78.0), IT (83.0), KE (83.7) and SF (74.5).

@inproceedings{bib_Visi_2022, AUTHOR = {Reddy, Oota Subba and Arora, Jashn and ROWTULA, VIJAY BAPANAIAH and Gupta, Manish and Surampudi, Bapiraju }, TITLE = {Visio-Linguistic Brain Encoding}, BOOKTITLE = {International Conference on Computational Linguistics}. YEAR = {2022}}

Enabling effective brain-computer interfaces requires understanding how the human brain encodes stimuli across modalities such as visual, language (or text), etc. Brain encoding aims at constructing fMRI brain activity given a stimulus. There exists a plethora of neural encoding models which study brain encoding for single mode stimuli: visual (pretrained CNNs) or text (pretrained language models). Few recent papers have also obtained separate visual and text representation models and performed late-fusion using simple heuristics. However, previous work has failed to explore: (a) the effectiveness of image Transformer models for encoding visual stimuli, and (b) co-attentive multi-modal modeling for visual and text reasoning. In this paper, we systematically explore the efficacy of image Transformers (ViT, DEiT, and BEiT) and multi-modal Transformers (VisualBERT, LXMERT, and CLIP) for brain encoding. Extensive experiments on two popular datasets, BOLD5000 and Pereira, provide the following insights. (1) To the best of our knowledge, we are the first to investigate the effectiveness of image and multi-modal Transformers for brain encoding. (2) We find that VisualBERT, a multi-modal Transformer, significantly outperforms previously proposed single-mode CNNs, image Transformers as well as other previously proposed multi-modal models, thereby establishing new state-of-theart. The supremacy of visio-linguistic models raises the question of whether the responses elicited in the visual regions are affected implicitly by linguistic processing even when passively viewing images. Future fMRI tasks can verify this computational insight in an appropriate experimental setting.

@inproceedings{bib_A_Co_2022, AUTHOR = {KAUSHIK, PRAMOD SIVARAM and Kummetha, Sneha and Moodley, Perusha and Surampudi, Bapiraju }, TITLE = {A Conservative Q-Learning approach for handling distribution shift in sepsis treatment strategies}, BOOKTITLE = {Technical Report}. YEAR = {2022}}

Sepsis is a leading cause of mortality and its treatment is very expensive. Sepsis treatment is also very challenging because there is no consensus on what interventions work best and different patients respond very differently to the same treatment. Deep Reinforcement Learning methods can be used to come up with optimal policies for treatment strategies mirroring physician actions. In the healthcare scenario, the available data is mostly collected offline with no interaction with the environment, which necessitates the use of offline RL techniques. The Offline RL paradigm suffers from action distribution shifts which in turn negatively affects learning an optimal policy for the treatment. In this work, a Conservative-Q Learning (CQL) algorithm is used to mitigate this shift and its corresponding policy reaches closer to the physicians policy than conventional deep Q Learning. The policy learned could help clinicians in Intensive Care Units to make better decisions while treating septic patients and improve survival rate.

@inproceedings{bib_Brai_2021, AUTHOR = {Qutubuddin, Md and Gibo, Tilahun Kochito and Surampudi, Bapiraju and Narri, Yadaiah }, TITLE = {Brain Affective System Inspired Control Architecture: An Application to Nonlinear System}, BOOKTITLE = {IEEE Access}. YEAR = {2021}}

This paper presents a bio-inspired intelligent controller to enhance the performance of a nonlinear system. The controller is designed by capturing the emotional intelligence of mammalian brain mediated by the limbic system in which certain parts are responsible for generating emotions and these can be combined together as brain affective system inspired control architecture (BASIC). This controller has been used to cope with nonlinearities present in control applications. In this paper brain inspired control architecture is proposed while incorporating the sensory cortex explicitly in the architecture and the computational equations of various modules are suitably developed. The performance of the proposed controller is analyzed on a typical nonlinear system, namely, Permanent magnet synchronous motor for speed control, harmonics in stator phase currents and ripples produced in electromagnetic torque at different operating conditions. Its performance is compared with the existing control techniques in offline simulations as well as in real-time Hardware-in-loop environment. The results establish the computational efficiency, accuracy and robustness of the proposed controller.

@inproceedings{bib_What_2021, AUTHOR = {Injamuri, Krutika and Komanduri, Sai Somanath and Bhagvati, Chakravarthy and Surampudi, Bapiraju }, TITLE = {What Do RDMs Capture in Brain Responses and Computational Models?}, BOOKTITLE = {Machine Learning: Theoretical Foundations and Practical Applications}. YEAR = {2021}}

The question of how the human brain represents information, and if insights from this knowledge help us in formulating better deep learning algorithms, is still open. This paper attempts to address these questions by investigating the brain responses when participants viewed images of objects belonging to a variety of categories. We adopt the representational similarity analysis (RSA) framework wherein the brain responses are projected into a representational dissimilarity (RD) space so that responses from multiple subjects can be assessed on a common basis. We performed RSA of the brain responses of 15 participants on a set of 92 images (from the Algonauts challenge data set) Cichy et al. (The Algonauts project: a platform for communication between the sciences of biological and artificial intelligence 2019 [4]). The results reveal that human brain responses exhibit appropriate qualitative similarities observed in the image space. We then evaluated various deep networks such as VGGnet, AlexNet, and Inception in Siamese configuration with RD-loss function and identified hidden layers that had maximal similarity to human RD matrices (RDM). The best layers then were subjected to further analysis by looking at the image pairs corresponding to high and

@inproceedings{bib_ECG__2021, AUTHOR = {Annam, Jagadeeswara Rao and Sujatha, R. and Somala, Jayaprada and Prasad, Gutta V. S. N. R. V. and Satyala, Narayana and Surampudi, Bapiraju }, TITLE = {ECG Feature Extraction}, BOOKTITLE = {Innovative Trends in Computational Intelligence}. YEAR = {2021}}

ECG analysis comprises the following steps: preprocessing, segmentation, feature extraction, and classification of heart-beat instances to detect cardiac arrhythmias. This work focuses on the first three steps in cardiac arrhythmia analysis. Since no publicly available feature sets are available for the ECG arrhythmia detection problem, researchers have to explicitly extract the feature sets from the raw signals available in PhysioBank repository. In this work, the complete feature extraction pipeline has been developed using Matlab environment. Both the Matlab code and resulting feature sets for all records have been deposited and made publicly available in the Github repository. This is one of the major

@inproceedings{bib_Effe_2021, AUTHOR = {Dwivedi, Madhukar and Dubey, Neha and Pansari, Aditya Jain and Surampudi, Bapiraju and Das, Meghoranjani and Guha, Maushumi and Banerjee, Rahul and Pramanick, Gobinda and Basu, Jayanti }, TITLE = {Effects of Meditation on Structural Changes of the Brain in Patients With Mild Cognitive Impairment or Alzheimer’s Disease Dementia}, BOOKTITLE = {Frontiers in Human Neuroscience}. YEAR = {2021}}

long-term meditation intervention. MCI or mild AD patients underwent detailed clinical and neuropsychological assessment and were assigned into meditation or non-meditation groups. High resolution T1-weighted magnetic resonance images (MRI) were acquired at baseline and after 6 months. Longitudinal symmetrized percentage changes (SPC) in cortical thickness and gray matter volume were estimated. Left caudal middle frontal, left rostral middle frontal, left superior parietal, right lateral orbitofrontal, and right superior frontal cortices showed changes in both cortical thickness and gray matter volume; the left paracentral cortex showed changes in cortical thickness; the left lateral occipital, left superior frontal, left banks of the superior temporal sulcus (bankssts), and left medial orbitofrontal cortices showed changes in gray matter volume. All these areas exhibited significantly higher SPC values in meditators as compared to non-meditators. Conversely, the left lateral occipital, and right posterior cingulate cortices showed significantly lower SPC values for cortical thickness in the meditators. In hippocampal subfields analysis, we observed significantly higher SPC in gray matter volume of the left CA1, molecular layer HP, and CA3 with a trend for increased gray matter volume in most other areas. No significant changes were found for the hippocampal

@inproceedings{bib_IMLE_2021, AUTHOR = {Reddy, Likith and Talwar, Vivek and Alle, Shanmukh and Surampudi, Bapiraju and U, Deva Priyakumar }, TITLE = {IMLE-Net: An Interpretable Multi-level Multi-channel Model for ECG Classification}, BOOKTITLE = {International Conference on Systems, Man, and Cybernetics}. YEAR = {2021}}

Early detection of cardiovascular diseases is crucial for effective treatment and an electrocardiogram (ECG) is pivotal for diagnosis. The accuracy of Deep Learning based methods for ECG signal classification has progressed in recent years to reach cardiologist-level performance. In clinical settings, a cardiologist makes a diagnosis based on the standard 12-channel ECG recording. Automatic analysis of ECG recordings from a multiple-channel perspective has not been given enough attention, so it is essential to analyze an ECG recording from a multiple-channel perspective. We propose a model that leverages the multiple-channel information available in the standard 12-channel ECG recordings and learns patterns at the beat, rhythm, and channel level. The experimental results show that our model achieved a macro-averaged ROC-AUC score of 0.9216, mean accuracy of 88.85% and a maximum F1 score of 0.8057 on the PTB-XL dataset. The attention visualization results from the interpretable model are compared against the cardiologist’s guidelines to validate the correctness and usability

@inproceedings{bib_Atte_2021, AUTHOR = {SHUKLA, ANUJ KUMAR and Surampudi, Bapiraju }, TITLE = {Attention mediates the influence of numerical magnitude on temporal processing}, BOOKTITLE = {Scientific Reports}. YEAR = {2021}}

The processing of time and numbers has been fundamental to human cognition. One of the prominent theories of magnitude processing, a theory of magnitude (ATOM), suggests that a generalized magnitude system processes space, time, and numbers; thereby, the magnitude dimensions could potentially interact with one another. However, more recent studies have found support for domain-specific magnitude processing and argued that the magnitudes related to time and number are processed through distinct mechanisms. Such mixed findings have raised questions about whether these magnitudes are processed independently or share a common processing mechanism. In the present study, we examine the influence of numerical magnitude on temporal processing. To investigate, we conducted two experiments using a temporal comparison task, wherein we presented positive and negative numerical …

@inproceedings{bib_Moto_2021, AUTHOR = {Bera, Krishn and SHUKLA, ANUJ KUMAR and Surampudi, Bapiraju }, TITLE = {Motor Chunking in Internally-guided Sequence Learning}, BOOKTITLE = {Annual Conference of the Cognitive Science Societ}. YEAR = {2021}}

The studies have shown that with practice, the sequential motor movements are executed as a group of sub-sequences called motor chunks [Lashley, 1951; Rosenbaum et al., 1983, Sakai et al. 2003]. The temporal patterns of these sequential movements reflect a hierarchical and structured representation of the motor sequence. Previous studies [Bapi et al. 2000, Verwey et al. 2003, Sakai et al. 2003] have used multiple canonical paradigms such as serial reaction time task, discrete sequence production task and m×n task to investigate chunking in externally-guided sequencing tasks, i.e., tasks where the external (visual) stimuli explicitly specify the sequence of motor actions. We argue that real-life, practical motor tasks such as playing a keyboard or solving a Rubik’s cube involve internally-guided motor actions, which are volitionally planned and are not externally-specified by the environment. Not many previous studies have examined the chunking phenomenon in such tasks. Motivated by this, our present study investigates chunking in internally-guided discrete sequencing.

@inproceedings{bib_Moto_2021, AUTHOR = {Bera, Krishn and SHUKLA, ANUJ KUMAR and Surampudi, Bapiraju }, TITLE = {Motor Chunking in Internally Guided Sequencing}, BOOKTITLE = {Brain Sciences}. YEAR = {2021}}

Motor skill learning involves the acquisition of sequential motor movements with practice. Studies have shown that we learn to execute these sequences efficiently by chaining several elementary actions in sub-sequences called motor chunks. Several experimental paradigms, such as serial reaction task, discrete sequence production, and m × n task, have investigated motor chunking in externally specified sequencing where the environment or task paradigm provides the sequence of stimuli, i.e., the responses are stimulus driven. In this study, we examine motor chunking in a class of more realistic motor tasks that involve internally guided sequencing where the sequence of motor actions is self-generated or internally specified. We employ a grid-navigation task as an exemplar of internally guided sequencing to investigate practice-driven performance improvements due to motor chunking. The participants performed the grid-sailing task (GST) (Fermin et al., 2010), which required navigating (by executing sequential keypresses) a 10 × 10 grid from start to goal position while using a particular type of key mapping between the three cursor movement directions and the three keyboard buttons. We provide empirical evidence for motor chunking in grid-navigation tasks by showing the emergence of subject-specific, unique temporal patterns in response times. Our findings show spontaneous chunking without pre-specified or externally guided structures while replicating the earlier results with a less constrained, internally guided sequencing paradigm

@inproceedings{bib_Cogn_2021, AUTHOR = {Bera, Krishn and SHUKLA, ANUJ KUMAR and Surampudi, Bapiraju }, TITLE = {Cognitive and Motor Learning in Internally-Guided Motor Skills}, BOOKTITLE = {Frontiers in psychology}. YEAR = {2021}}

Our everyday experiences are an excellent demonstration of the surprisingly adaptive and fluid learning behavior that is orchestrated by the human brain. Such a learning behavior is a hallmark of human cognitive ability and spans a broad spectrum of tasks. Ranging from complex tasks such as cycling and driving to seemingly simpler ones such as typing and grasping movements, all tasks involve the acquisition of skillful behavior. Skill learning is a natural behavioral phenomenon concerned with the acquisition of the ability to perform tasks proficiently. Motor skill learning refers to learning a specific subclass of skills that involve sequential motor movements such that they are executed accurately and quickly with practice (Newell, 1991; Clegg et al., 1998; Haibach et al., 2017; Schmidt et al., 2019). Much of the early interest in motor sequencing focused on investigating the typical behavioral phenomenon in sequence learning tasks (Lashley, 1951; Hebb, 1961; Fitts and Posner, 1967). This has led to the formulation of many serial order canonical experimental tasks such as the m × n task (Hikosaka et al., 1995; Bapi et al., 2000, 2006) and discrete sequence production (DSP) task (Verwey, 2001; Abrahamse et al., 2013; Verwey et al., 2015) in the explicit domain and serial reaction time (SRT) task (Nissen and Bullemer, 1987; Willingham, 1999; Robertson, 2007) in the implicit domain. While explicit learning involves conscious awareness of what is being learned, implicit learning occurs without conscious awareness of learning. Subsequent research has extensively used these paradigms to understand the brain processes involved in sequence learning, memory, attention, etc.

@inproceedings{bib_Nume_2021, AUTHOR = {Shukla, Anuj and Surampudi, Bapiraju }, TITLE = {Numerical Magnitude Affects Accuracy but Not Precision of Temporal Judgments}, BOOKTITLE = {Frontiers in Human Neuroscience}. YEAR = {2021}}

A Theory of Magnitude (ATOM) suggests that space, time, and quantities are processed through a generalized magnitude system. ATOM posits that task-irrelevant magnitudes interfere with the processing of task-relevant magnitudes as all the magnitudes are processed by a common system. Many behavioural and neuroimaging studies have found support in favour of a common magnitude processing system. However, it is largely unknown whether such cross-domain monotonic mapping arises from a change in accuracy of the magnitude judgments or results from changes in precision of the processing of magnitude. Therefore, in the present study, we examined whether large numerical magnitude affects temporal accuracy or temporal precision or both. In other words, whether numerical magnitudes change our temporal experience or simply bias our duration judgments. The temporal discrimination (between …

@inproceedings{bib_Mood_2020, AUTHOR = {Steephen, John Eric and Kummetha, Sneha and Obbineni, Siva Charan and Surampudi, Bapiraju }, TITLE = {Mood-congruent biases in facial emotion perception and their gender dependence}, BOOKTITLE = {International journal of psychology}. YEAR = {2020}}

This work was supported by the Science and Engineering Research Board, Department of Science and Technology, Ministry of Science and Technology, India, under Grant SR/CSI/PDF-21/2013. John Eric Steephen: Conception and design, analysis and interpretation of data, drafting the article. Sneha Kummetha: Design, collection and analysis of data. Siva Charan Obbineni: Design, collection and analysis of data. Raju Surampudi Bapi: Design, drafting the article.

@inproceedings{bib_Valu_2020, AUTHOR = {Bera, Krishn and Mandilwar, Yash and Surampudi, Bapiraju }, TITLE = {Value-of-Information based Arbitration between Model-based and Model-freeControl}, BOOKTITLE = {Annual meeting of the Cognitive Science Society.}. YEAR = {2020}}

There have been numerous attempts in explaining the general learning behaviours using model-based and model-free methods. While the model-based control is flexible yet computationally expensive in planning, the model-free control is quick but inflexible. Multiple arbitration schemes have been suggested to achieve the data efficiency and computational efficiency of model-based and model-free control schemes, respectively. In this context, we propose a quantitative ’valueof-information’ based arbitration between both the controllers in order to establish a general computational framework for skill learning. The interacting model-based and model-free reinforcement learning processes are arbitrated using an uncertainty-based value-of-information estimation. We further show that our algorithm performs better than Q-learning as well as Q-learning with experience replay.

@inproceedings{bib_Grid_2020, AUTHOR = {Bera, Krishn and SHUKLA, ANUJ KUMAR and Surampudi, Bapiraju }, TITLE = {Grid-Navigation Tasks involve Skill Learning}, BOOKTITLE = {Annual meeting of the Cognitive Science Society.}. YEAR = {2020}}

Several canonical experimental paradigms (serial reaction task, mxn task, etc.) have been proposed to study the typical behavioural phenomenon in sequential key-press tasks. However, not much work has been done on studying motor sequencing in grid-navigation tasks. In this work, using empirical examinations, we systematically show grid-navigation task as an instance of skill learning paradigm. The participants performed Grid-Sailing Task (GST), which required navigating (by executing sequential key-presses) a 5x5 grid from start to goal position while using a particular key-mapping among the 3 cursor movement directions and the 3 keyboard buttons. We employ two different experiments to argue for the learning of cognitive strategies as well as motor sequences. By rejecting the motor adaptation argument and validating the law of practice, we characterize GST as a skill learning task. We further argue for advantages of GST as a general, canonical task over others for use in skill learning studies.

@inproceedings{bib_MOTO_2020, AUTHOR = {Bera, Krishn and SHUKLA, ANUJ KUMAR and Surampudi, Bapiraju }, TITLE = {MOTOR CHUNKING DURING SEQUENCE LEARNING IN GRID-NAVIGATION TASKS}, BOOKTITLE = {Annual meeting of the Cognitive Science Society.}. YEAR = {2020}}

Several canonical experimental paradigms (serial reaction task, mxn task, etc.) have been proposed to study the typical behavioural phenomena in a sequential motor key-press task. The repeated execution of visuomotor sequences in such paradigms lead to overall performance improvement such that the inter-response intervals in between certain sub-sequences decreases as compared to that across other sub-sequences. This efficient and hierarchical cluster organisation is called \textit{motor chunking}. We provide empirical evidence for motor chunking in grid-navigation sequencing tasks. The participants performed Grid-Sailing Task (GST) [Fermin et. al., 2010] that required navigating a 10x10 grid from start to goal position while using a particular key-mapping between the 3 cursor movement directions and the 3 keyboard buttons. This study confirms the emergence of subject-specific, unique temporal patterns related to chunking after substantial practice.

@inproceedings{bib_Clas_2020, AUTHOR = {Annam, Jagadeeswara Rao and K, Srinivas and Ch, Sureshbabu and Somala, Jayaprada and Surampudi, Bapiraju }, TITLE = {Classification of ECG Heartbeat Arrhythmia: A Review}, BOOKTITLE = {Procedia Computer Science}. YEAR = {2020}}

Manual identification of ECG heart-beat classes by cardiologists is time consuming and cumbersome. These professionals rely on computer based methods for determination of these heart-disease types. In this work, existing literature is organized into a proposed taxonomy based on dichotomies involving full time series-based versus feature-based, AAMI versus Non-AAMI, and inter-patient versus intra-patient based distinctions. The basic contributions of this work are systematic review of literature on heart-beat abnormality detection, identifying research gaps and the research issues unmet sofar in the literature to propose novel approaches for addressing these gaps.

@inproceedings{bib_Mode_2019, AUTHOR = {D, AMARNATH and Surampudi, Bapiraju }, TITLE = {Modelling the Influence of Affect on Cognitive Processing using Chrest and Nengo }, BOOKTITLE = {International Conference on Cognitive Modeling}. YEAR = {2019}}

It has long been understood that there is an interplay between affect and cognition((Kort & Reilly, 2003), but this interaction, based on the recent chess studies((Guntz, Crowley, Vaufreydaz, Balzarini, & Dessus, 2018), is much more intertwined than what the established theories postulate. To understand the underlying mechanisms in greater detail we propose an integrated model using Chrest and Nengo. We analyze the results based on simulations with data from previous empirical studies. Keywords: Affect, Cognition , Chess, Nengo , Chrest

@inproceedings{bib_Valu_2019, AUTHOR = {Bera, Krishn and Mandilwar, Yash and Surampudi, Bapiraju }, TITLE = {Value-of-Information based Arbitration between Model-based anModel-free Control}, BOOKTITLE = {Technical Report}. YEAR = {2019}}

There have been numerous attempts in explaining the general learning behaviours using model-based and modelfree methods. While the model-based control is flexible yet computationally expensive in planning, the model-free control is quick but inflexible. The model-based control is therefore immune from reward devaluation and contingency degradation. Multiple arbitration schemes have been suggested to achieve the data efficiency and computational efficiency of modelbased and model-free control respectively. In this context, we propose a quantitative ’value of information’ based arbitration between both the controllers in order to establish a general computational framework for skill learning. The interacting model-based and model-free reinforcement learning processes are arbitrated using an uncertainty-based value of information. We further show that our algorithm performs better than Qlearning as well as Q-learning with experience replay

@inproceedings{bib_A_Fr_2019, AUTHOR = {D, AMARNATH and Surampudi, Bapiraju }, TITLE = {A Framework for Affective and Sustainable Learning}, BOOKTITLE = {International Journal of Affective Engineering}. YEAR = {2019}}

@inproceedings{bib_Atyp_2019, AUTHOR = {HARLALKA, VATIKA and Surampudi, Bapiraju and Krishnanunni, Vinod Palakkad and Roy, Dipanjan }, TITLE = {Atypical Flexibility in Dynamic Functional Connectivity Quantifies the Severity in Autism Spectrum Disorder}, BOOKTITLE = {Frontiers Human Neuroscience}. YEAR = {2019}}

Resting-state functional connectivity (FC) analyses have shown atypical connectivity in autism spectrum disorder (ASD) as compared to typically developing (TD). However, this view emerges from investigating static FC overlooking the whole brain transient connectivity patterns. In our study, we investigated how age and disease influence the dynamic changes in functional connectivity of TD and ASD. We used resting-state functional magnetic resonance imaging (rs-fMRI) data stratified into three cohorts: children (7–11 years), adolescents (12–17 years), and adults (18+ years) for the analysis. The dynamic variability in the connection strength and the modular organization in terms of measures such as flexiblity, cohesion strength, and disjointness were explored for each subject to characterize the differences between ASD and TD. In ASD, we observed significantly higher inter-subject dynamic variability in connection strength as compared to TD. This hyper-variability relates to the symptom severity in ASD. We also found that whole-brain flexibility correlates with static modularity only in TD. Further, we observed a core-periphery organization in the resting-state, with Sensorimotor and Visual regions in the rigid core; and DMN and attention areas in the flexible periphery. TD also develops a more cohesive organization of sensorimotor areas. However, in ASD we found a strong positive correlation of symptom severity with flexibility of rigid areas and with disjointness of sensorimotor areas. The regions of the brain showing high predictive power of symptom severity were distributed across the cortex, with stronger bearings in the frontal, motor, and occipital cortices. Our study demonstrates that the dynamic framework best characterizes the variability in ASD.

@inproceedings{bib_Step_2019, AUTHOR = {REDDY, OOTA SUBBA and ROWTULA, VIJAY BAPANAIAH and Gupta, Manish and Surampudi, Bapiraju }, TITLE = {StepEncog: A Convolutional LSTM Autoencoder for Near-Perfect fMRI Encoding}, BOOKTITLE = {International Joint Conference on Neural Networks}. YEAR = {2019}}

Learning a forward mapping that relates stimuli to the corresponding brain activation measured by functional magnetic resonance imaging (fMRI) is termed as estimating encoding models. Computational tractability usually forces current encoding as well as decoding solutions to typically consider only a small subset of voxels from the actual 3D volume of activation. Further, while reconstructing stimulus information from brain activation (brain decoding) has received wider attention, there have been only a few attempts at constructing encoding solutions in the extant neuro-imaging literature. In this paper, we present StepEncog, a convolutional LSTM autoencoder model trained on fMRI voxels. The model can predict the entire brain volume rather than a small subset of voxels, as presented in earlier research works. We argue that the resulting solution avoids the problem of devising encoding models based on a rulebased selection of informative voxels and the concomitant issue of wide spatial variability of such voxels across participants. The perturbation experiments indicate that the proposed deep encoder indeed learns to predict brain activations with high spatial accuracy. On challenging universal decoder imaging datasets, our model yielded encouraging results.

@inproceedings{bib_Expe_2019, AUTHOR = {REDDY, OOTA SUBBA and Manwani, Naresh and Surampudi, Bapiraju }, TITLE = {Expert2Coder: Capturing Divergent Brain Regions Using Mixture of Regression Experts}, BOOKTITLE = {Technical Report}. YEAR = {2019}}

fMRI semantic category understanding using linguistic encoding models attempts to learn a forward mapping that relates stimuli to the corresponding brain activation. Stateof-the-art encoding models use a single global model (linear or non-linear) to predict brain activation (all the voxels) given the stimulus. However, the critical assumption in these methods is that a priori different brain regions respond the same way to all the stimuli, that is, there is no modularity or specialization assumed for any region. This goes against the modularity theory, supported by many cognitive neuroscience investigations suggesting that there are functionally specialized regions in the brain. In this paper we achieve this by clustering similar regions together and for every cluster we learn a different linear regression model using a mixture of linear experts model. The key idea here is that each linear expert captures the behaviour of similar brain regions. Given a new stimulus, the utility of the proposed model is twofold (i) predicts the brain activation as a weighted linear combination of the activations of multiple linear experts and (ii) to learn multiple experts corresponding to different brain regions. We argue that each expert captures activity patterns related to a particular region of interest (ROI) in the human brain. This study helps in understanding the brain regions that are activated together given different kinds of stimuli. Importantly, we suggest that the mixture of regression experts (MoRE) framework successfully combines the two principles of organization of function in the brain, namely that of specialization and integration. Experiments on fMRI data from paradigm 1 [1] where participants view linguistic stimuli show that the proposed MoRE model has better prediction accuracy compared to that of conventional models. Our model achieves a mean absolute error (MAE) of 3.94, with an R 2-score of 0.45 on this data set. This is an improvement over performance of traditional methods including, ridge regression (5.58 MAE, 0.15 R 2 -score), MLP (4.63 MAE, 0.35 R 2 -score). We also elaborate on the specializations captured by various experts in our mixture model and their implications.

@inproceedings{bib_A_Co_2018, AUTHOR = {Bera, Krishn and PARSOTTAM, SAVALIA TEJAS and Surampudi, Bapiraju }, TITLE = {A Computational Framework for Motor Skill Learning}, BOOKTITLE = {Technical Report}. YEAR = {2018}}

There have been numerous attempts in explaining the general learning behaviours by various cognitive models. Multiple hypotheses have been put further to qualitatively argue the best-fit model for motor skill acquisition task and its variations. In this context, for a discrete sequence production (DSP) task, one of the most insightful models is Verweys Dual Processor Model (DPM). It largely explains the learning and behavioural phenomenon of skilled discrete key-press sequences without providing any concrete computational basis of reinforcement. Therefore, we propose a quantitative explanation for Verweys DPM hypothesis by experimentally establishing a general computational framework for motor skill learning. We attempt combining the qualitative and quantitative theories based on a best-fit model of the experimental simulations of variations of dual processor models. The fundamental premise of sequential decision making for skill learning is based on interacting model-based (MB) and model-free (MF) reinforcement learning (RL) processes. Our unifying framework shows the proposed idea agrees well to Verweys DPM and Fitts three phases of skill learning. The accuracy of our model can further be validated by its statistical fit with the human-generated data on simple environment tasks like the grid-world. There have been numerous attempts in explaining the general learning behaviours by various cognitive models. Multiple hypotheses have been put further to qualitatively argue the best-fit model for motor skill acquisition task and its variations. In this context, for a discrete sequence production (DSP) task, one of the most insightful models is Verweys Dual Processor Model (DPM). It largely explains the learning and behavioural phenomenon of skilled discrete key-press sequences without providing any concrete computational basis of reinforcement. Therefore, we propose a quantitative explanation for Verweys DPM hypothesis by experimentally establishing a general computational framework for motor skill learning. We attempt combining the qualitative and quantitative theories based on a best-fit model of the experimental simulations of variations of dual processor models. The fundamental premise of sequential decision making for skill learning is based on interacting model-based (MB) and model-free (MF) reinforcement learning (RL) processes. Our unifying framework shows the proposed idea agrees well to Verweys DPM and Fitts three phases of skill learning. The accuracy of our model can further be validated by its statistical fit with the human-generated data on simple environment tasks like the grid-world.

@inproceedings{bib_An_A_2018, AUTHOR = {Steephen, John Eric and Obbineni, Siva C. and SNEHA, KUMMETHA and Surampudi, Bapiraju }, TITLE = {An Affective Adaptation Model Explaining the Intensity-Duration Relationship of Emotion}, BOOKTITLE = {Transactions on Affective Computing}. YEAR = {2018}}

Intensity and duration are both pertinent aspects of an emotional experience, yet how they are related is unclear. Though stronger emotions usually last longer, sometimes they abate faster than the weaker ones. We present a quantitative model of affective adaptation, the process by which emotional responses to unchanging affective stimuli weaken with time, that addresses this intensity-duration problem. The model, described by three simple linear algebraic equations, assumes that the relationship between an affective stimulus and its experiencer can be broken down into three parameters. Self-relevance and explanation level combine multiplicatively to determine emotion intensity whereas the interaction of these with explanatory ease determines its duration. The model makes predictions, consistent with available empirical data, about emotion intensity, its duration, and adaptation speed for different scenarios. It predicts when the intensity-duration correlation is positive, negative or even absent, thus offering a solution to the intensity-duration problem. The model also addresses the shortcomings of past models of affective adaptation with its enhanced predictive power and by offering a more complete explanation to empirical observations that earlier models explain inadequately or fail to explain altogether. The model has potential applications in areas such as virtual reality training, games, human-computer interactions, and robotics.

@inproceedings{bib_Crea_2018, AUTHOR = {D, AMARNATH and Surampudi, Bapiraju }, TITLE = {Creating an affordable ,effective, adaptive & personalized attention tasks for children with developmental disorders.}, BOOKTITLE = {Cognitive Science}. YEAR = {2018}}

The main challenge in studying cognition & designing effective tasks for children with learning disorders is creating personalized & adaptive tasks in line with the current abilities & mood of the child. The current study confronts this challenge by testing a new paradigm to access the current state of mind and adapting the tasks based on the current mood & abilities of the child. Children were given chess puzzles with various levels of difficulty (from just identifying the pieces, legal moves and eventually even capturing pieces with depth=1). while the children were performing the tasks the pupil-metric data (for cognitive load), facial expressions and the head pose were used to gauge the current-state and adapt the puzzles accordingly. Further development of dynamic feedback and providing rewards for looking at the right squares are also underway. custom software with off the shelf web-cameras were used as the current solutions in the market are prohibitively expensive for testing on large scale.

@inproceedings{bib_Holi_2018, AUTHOR = {D, AMARNATH and Surampudi, Bapiraju }, TITLE = {Holistic Framework for Accelerated Learning by Adapting and Personalizing Lesson Plan for Children based on Emotions}, BOOKTITLE = {International Symposium on Affective Science and Engineering}. YEAR = {2018}}

Affective education is an formal curriculum designed to help children better understand their feelings and respond to challenging situations,thereby, transforming themselves and the world around them.Emotions impact the learning ability at multiple levels(Attention,Memory and decision making etc.).Though they have been advancements in terms of the content(Rich multimedia-based lessons etc.) for effective learning,proportionate advancements in have not taken place in the affective learning domain-for example "How to adapt the learning based in the current situation?".Can we mitigate the adverse effects of emotions?This problem of learning is especially compounded for university students where each student has flood of information to absorb & assimilate and is constantly under under stress,furthermore,Personalized & Real time/continual mentoring by the teachers to students is not practical.We achieve this by capturing the real time gestures and facial expressions(based on universal facial expressions of 7 emotions The Task(Chess Puzzles) given to validate the effectiveness of Methods show significant Improvement on sample size of 80 students

@inproceedings{bib_Mult_2018, AUTHOR = {SURAMPUDI, SRINIWAS GOVINDA and GIRISHKUMAR, NAIK SHRUTI and Surampudi, Bapiraju and Jirsa, Viktor K. and Sharma, Avinash and Roy, Dipanjan }, TITLE = {Multiple Kernel Learning Model for Relating Structural and Functional Connectivity in the Brain}, BOOKTITLE = {Scientific Reports}. YEAR = {2018}}

A challenging problem in cognitive neuroscience is to relate the structural connectivity (SC) to the functional connectivity (FC) to better understand how large-scale network dynamics underlying human cognition emerges from the relatively fxed SC architecture. Recent modeling attempts point to the possibility of a single difusion kernel giving a good estimate of the FC. We highlight the shortcomings of the single-difusion-kernel model (SDK) and propose a multi-scale difusion scheme. Our multiscale model is formulated as a reaction-difusion system giving rise to spatio-temporal patterns on a fxed topology. We hypothesize the presence of inter-regional co-activations (latent parameters) that combine difusion kernels at multiple scales to characterize how FC could arise from SC. We formulated a multiple kernel learning (MKL) scheme to estimate the latent parameters from training data. Our model is analytically tractable and complex enough to capture the details of the underlying biological phenomena. The parameters learned by the MKL model lead to highly accurate predictions of subject-specifc FCs from test datasets at a rate of 71%, surpassing the performance of the existing linear and non-linear models. We provide an example of how these latent parameters could be used to characterize age-specifc reorganization in the brain structure and function.

@inproceedings{bib_Two__2018, AUTHOR = {, Alok Singh and Alshamir, Abobakr Khalil and Surampudi, Bapiraju }, TITLE = {Two swarm intelligence approaches for tuning extreme learning machine}, BOOKTITLE = {International Journal of Machine Learning and Cybernetics}. YEAR = {2018}}

Extreme learning machine (ELM) is a new algorithm for training single-hidden layer feedforward neural networks which provides good performance as well as fast learning speed. ELM tends to produce good generalization performance with large number of hidden neurons as the input weights and hidden neurons biases are randomly initialized and remain unchanged during the learning process, and the output weights are analytically determined. In this paper, two swarm intelligence based metaheuristic techniques, viz. Artificial Bee Colony (ABC) and Invasive Weed Optimization (IWO) are proposed for tuning the input weights and hidden biases. The proposed approaches are called ABC-ELM and IWO-ELM in which the input weights and hidden biases are selected using ABC and IWO respectively and the output weights are computed using the Moore-Penrose (MP) generalized inverse. The proposed approaches are tested on different benchmark classification data sets and simulations show that the proposed approaches obtain good generalization performance in comparison to the other techniques available in the literature.

@inproceedings{bib_Mach_2018, AUTHOR = {SINGH, NOOR PRATAP and Surampudi, Bapiraju and Krishnanunni, Vinod Palakkad }, TITLE = {Machine learning models to predict the progression from early to late stages of papillary renal cell carcinoma}, BOOKTITLE = {Computers in Biology and Medicine}. YEAR = {2018}}

Papillary Renal Cell Carcinoma (PRCC) is a heterogeneous disease with variations in disease progression and clinical outcomes. The advent of next generation sequencing techniques (NGS) has generated data from patients that can be analysed to develop a predictive model. In this study, we have adopted a machine learning approach to identify biomarkers and build classifiers to discriminate between early and late stages of PRCC from gene expression profiles. A machine learning pipeline incorporating different feature selection algorithms and classification models is developed to analyse RNA sequencing dataset (RNASeq). Further, to get a reliable feature set, we extracted features from different partitions of the training dataset and aggregated them into feature sets for classification. We evaluated the performance of different algorithms on the basis of 10-fold cross validation and independent test dataset. 10-fold cross validation was also performed on a microarray dataset of PRCC. A random forest based feature selection (varSelRF) yielded minimum number of features (104) and a best performance with area under Precision Recall curve (PR-AUC) of 0.804, MCC (Matthews Correlation Coefficient) of 0.711 and accuracy of 88% with Shrunken Centroid classifier on a test dataset. We identified 80 genes that are consistently altered between stages by different feature selection lgorithms. The extracted features are related to cellular components - centromere, kinetochore and spindle, and biological process mitotic cell cycle. These observations reveal potential mechanisms for an increase in chromosome instability in the late stage of PRCC. Our study demonstrates that the gene expression profiles can be used to classify stages of PRCC.

@inproceedings{bib_Age,_2018, AUTHOR = {HARLALKA, VATIKA and Surampudi, Bapiraju and Krishnanunni, Vinod Palakkad and Roy, Dipanjan }, TITLE = {Age, Disease, and Their Interaction Effects on Intrinsic Connectivity of Children and Adolescents in Autism Spectrum Disorder Using Functional Connectomics}, BOOKTITLE = {Brain Connectivity}. YEAR = {2018}}

Brain connectivity analysis has provided crucial insights to pinpoint the differences between autistic and typically developing (TD) children during development. The aim of this study is to investigate the functional connectomics of autism spectrum disorder (ASD) versus TD and underpin the effects of development, disease, and their interactions on the observed atypical brain connectivity patterns. Resting-state functional magnetic resonance imaging (rs-fMRI) from the Autism Brain Imaging Data Exchange (ABIDE) data set, which is stratified into two cohorts: children (9–12 years) and adolescents (13–16 years), is used for the analysis. Differences in various graph theoretical network measures are calculated between ASD and TD in each group. Furthermore, two-factor analysis of variance test is used to study the effect of age, disease, and their interaction on the network measures and the network edges. Furthermore, the differences in connection strength between TD and ASD subjects are assessed using network-based statistics. The results showed that ASD exhibits increased functional integration at the expense of decreased functional segregation. In ASD adolescents, there is a significant decrease in odularity suggesting a less robust modular organization, and an increase in participation coefficient suggesting more random integration and widely distributed connection edges. Furthermore, there is significant hypoconnectivity observed in the adolescent group especially in the default mode network, while the children group shows both hyper- and hypoconnectivity. This study lends support to a model of global atypical connections and further identifies functional networks and areas that are independently affected by age, disease, and their interaction.

@inproceedings{bib_Expe_2018, AUTHOR = {SNEHA, KUMMETHA and KAUSHIK, PRAMOD SIVARAM and SHUKLA, ANUJ KUMAR and Surampudi, Bapiraju }, TITLE = {Experimental And Computational Investigation of the Effects of Variable RSI on Sequence Learning}, BOOKTITLE = {Conference on Cognitive Computational Neuroscience}. YEAR = {2018}}

In this study, we investigated the effects of variable Response-to-Stimulus interval (RSI) on sequence learning using both empirical and computational methods. In the empirical study, the serial reaction time task (SRT) was conducted which was followed by free generation and recognition tasks. Results showed that learning becomes explicit with increase in RSI despite its varying temporal nature. We constructed a computational model based on modified Elman network architecture to obtain a functional account of the empirical findings. The model illustrates how explicit learning could emerge due to a longer temporal window between stimuli which could potentially give insights into the mechanisms of sequence learning in variable RSI conditions.

@inproceedings{bib_A_De_2018, AUTHOR = {ROWTULA, VIJAY BAPANAIAH and REDDY, OOTA SUBBA and Gupta, Manish and Surampudi, Bapiraju }, TITLE = {A Deep Autoencoder for Near-Perfect fMRI Encoding}, BOOKTITLE = {Neural Information Processing Systems Workshops}. YEAR = {2018}}

Encoding models of functional magnetic resonance imaging (fMRI) data attempt to learn a forward mapping that relates stimuli to the corresponding brain activation. Computational tractability usually forces current encoding as well as decoding solutions to typically consider only a small subset of voxels from the actual 3D volume of activation. Further, while brain decoding has received wider attention, there have been only a few attempts at constructing encoding solutions in the extant neuroimaging literature. In this paper, we present a deep autoencoder consisting of convolutional neural networks in tandem with long short-term memory (CNNLSTM) model. The model is trained on fMRI slice sequences and predicts the entire brain volume rather than a small subset of voxels from the information in stimuli (text and image). We argue that the resulting solution avoids the problem of devising encoding models based on a rule-based selection of informative voxels and the concomitant issue of wide spatial variability of such voxels across participants. The perturbation experiments indicate that the proposed deep encoder indeed learns to predict brain activations with high spatial accuracy. On the challenging universal decoder imaging datasets (Pereira et al., 2018), our model yielded encouraging results.

@inproceedings{bib_fMRI_2018, AUTHOR = {REDDY, OOTA SUBBA and Manwani, Naresh and Surampudi, Bapiraju }, TITLE = {fMRI Semantic Category Decoding using Linguistic Encoding of Word Embeddings}, BOOKTITLE = {International Conference on Neural Information Processing}. YEAR = {2018}}

The dispute of how the human brain represents conceptual knowledge has been argued in many scientific fields. Brain imaging studies have shown that the spatial patterns of neural activation in the brain are correlated with thinking about different semantic categories of words (for example, tools, animals, and buildings) or when viewing the related pictures. In this paper, we present a computational model that learns to predict the neural activation captured in functional magnetic resonance imaging (fMRI) data of test words. Unlike the models with hand-crafted features that have been used in the literature, in this paper we propose a novel approach wherein decoding models are built with features extracted from popular linguistic encodings of Word2Vec, GloVe, Meta-Embeddings in conjunction with the empirical fMRI data associated with viewing several dozen concrete nouns. We compared these models with several other models that use word features extracted from FastText, Randomly-generated features, Mitchell’s 25 features [1]. The experimental results show that the predicted fMRI images using Meta-Embeddings meet the state-of-the-art performance. Although models with features from GloVe and Word2Vec predict fMRI images similar to the state-of-the-art model, model with features from Meta-Embeddings predicts significantly better. The proposed scheme that uses popular linguistic encoding offers a simple and easy approach for semantic decoding from fMRI experiments.

@inproceedings{bib_Mixt_2018, AUTHOR = {REDDY, OOTA SUBBA and AVVARU, ADITHYA and Manwani, Naresh and Surampudi, Bapiraju }, TITLE = {Mixture of Regression Experts in fMRI Encoding}, BOOKTITLE = {Neural Information Processing Systems Workshops}. YEAR = {2018}}

fMRI semantic category understanding using linguistic encoding models attempt to learn a forward mapping that relates stimuli to the corresponding brain activation. Classical encoding models use linear multi-variate methods to predict the brain activation (all voxels) given the stimulus. However, these methods essentially assume multiple regions as one large uniform region or several independent regions, ignoring connections among them. In this paper, we present a mixture of expertsbased model where a group of experts captures brain activity patterns related to particular regions of interest (ROI) and also show the discrimination across different experts. The model is trained word stimuli encoded as 25-dimensional feature vectors as input and the corresponding brain responses as output. Given a new word (25-dimensional feature vector), it predicts the entire brain activation as the linear combination of multiple experts’ brain activations. We argue that each expert learns a certain region of brain activations corresponding to its category of words, which solves the problem of identifying the regions with a simple encoding model. We showcase that proposed mixture of experts-based model indeed learns region-based experts to predict the brain activations with high spatial accuracy.

@inproceedings{bib_The__2017, AUTHOR = {Nallapu, Bhargav Teja and Surampudi, Bapiraju and Rougier, Nicolas P. }, TITLE = {The art of scaling up : a computational account on action selection in basal ganglia}, BOOKTITLE = {International Joint Conference on Neural Networks}. YEAR = {2017}}

What makes a computational neuronal model ’large scale’ ? Is it the number of neurons modeled? Or the number of brain regions modeled in a network? Most of the higher cognitive processes span across co-ordinated activity in a network of different brain areas. However at the same time, the basic information transfer takes place at a single neuron level, together with multiple other neurons. We explore modeling a neural system involving some areas of cortex, the basal ganglia (BG) and thalamus for the process of decision making, using a largescale neural engineering framework, Nengo. Early results tend to replicate the known neural activity patterns as found in the previous action selection model by Guthrie et al. 2013, besides operating with a larger neuronal populations. The power of converting algorithms to efficiently weighed neural networks in Nengo (Stewart et al. 2009 and Bekolay et al. 2013) is exploited in this work. Crucial aspects in a computational model, like parameter tuning and detailed neural implementations, while moving from a simplistic to large-scale model, are studied.

@inproceedings{bib_Meta_2017, AUTHOR = {GIRISHKUMAR, NAIK SHRUTI and Reddy, Oota Subba and Banerjee, Arpan and Roy, Dipanjan and Surampudi, Bapiraju }, TITLE = {Metastability of Cortical BOLD Signals in Maturation and Senescence }, BOOKTITLE = {International Joint Conference on Neural Networks}. YEAR = {2017}}

We assess change in metastability to characterize age-effects on the dynamic repertoire of the functional networks at rest. Resting state fMRI signals from each subject (N=48) have been used and metastability is evaluated as the standard deviation of mean phase synchrony of BOLD signals across whole-brain as well as across known resting state networks. The results suggest that significant whole-brain metastability changes occur between middle to old age. We also demonstrate that static time-averaged FC largely undermines age-effects on the interaction between functional networks. Discriminant Function Analysis reveals existence of two different patterns of change in metastability, which maximally discriminates between two different processes of maturation and ageing.

@inproceedings{bib_Big__2017, AUTHOR = {Sengupta, Rakesh and Surampudi, Bapiraju and Melcher, David }, TITLE = {Big and small numbers: Empirical support for a single, flexible mechanism for numerosity perception}, BOOKTITLE = {Attention Perception & Psychophysics}. YEAR = {2017}}

The existence of perceptually distinct numerosity ranges has been proposed for small (i.e., subitizing range) and larger numbers based on differences in precision, Weber fractions, and reaction times. This raises the question of whether such dissociations reflect distinct mechanisms operating across the two numerosity ranges. In the present work, we explore the predictions of a single-layer recurrent on-center, off-surround network model of attentional priority that has been applied to object individuation and enumeration. Activity from the network can be used to model various phenomena in the domain of visual number perception based on a single parameter: the strength of inhibition between nodes. Specifically, higher inhibition allows for precise representation of small numerosities, while low inhibition is preferred for high numerosities. The model makes novel predictions, including that enumeration of …

@inproceedings{bib_Expe_2017, AUTHOR = {SANKAR, REMYA and SHUKLA, ANUJ KUMAR and Surampudi, Bapiraju }, TITLE = {Experimental and Computational Investigation of the Effect of Caffeine on Human Time Perception}, BOOKTITLE = {Cognitive Science}. YEAR = {2017}}

Perception of time is an active process that takes place continually. However, we are yet to learn its exact mechanisms conclusively. The temporal bisection task is ideal to investigate the circuitry underlying time perception. Caffeine, a commonly used stimulant, has been known to play a role in modulation of time perception. The objective of this article is to explore the role of caffeine, a neuromodulator, in the perception of time in human beings by conducting suitable experiments. The experiment shows that an expansion of time is perceived by subjects after caffeine ingestion and that caffeine has an accelerating effect on our time perception system. Additionally, we present a preliminary 2-step decision model that fits the results of the experiment and potentially gives insights into the mechanisms of caffeine. We conclude by pointing out future directions towards a more biologically realistic computational model.

@inproceedings{bib_Bili_2017, AUTHOR = {Alladi, Suvarna and Bak, Thomas H and Shailaja, Mekala and Gollahalli, Divyaraj and Rajan, Amulya and Surampudi, Bapiraju and Michael, and Duggirala, Vasanta and Chaudhuri, Jaydip Ray }, TITLE = {Bilingualism delays the onset of behavioral but not aphasic forms of frontotemporal dementia}, BOOKTITLE = {Neuropsychologia}. YEAR = {2017}}

Bilingualism has been found to delay onset of dementia and this has been attributed to an advantage in executive control in bilinguals. However, the relationship between bilingualism and cognition is complex, with costs as well as benefits to language functions. To further explore the cognitive consequences of bilingualism, the study used Frontotemporal dementia (FTD) syndromes, to examine whether bilingualism modifies the age at onset of behavioral and language variants of Frontotemporal dementia (FTD) differently. Case records of 193 patients presenting with FTD (121 of them bilingual) were examined and the age at onset of the first symptoms were compared between monolinguals and bilinguals. A significant effect of bilingualism delaying the age at onset of dementia was found in behavioral variant FTD (5.7 years) but not in progressive nonfluent aphasia (0.7 years), semantic dementia (0.5 years …

@inproceedings{bib_The__2017, AUTHOR = {Nallapu, Bhargav Teja and Surampudi, Bapiraju and Rougier, Nicolas P. }, TITLE = {The art of scaling up: a computational account on action selection in basal ganglia}, BOOKTITLE = {International Joint Conference on Neural Networks}. YEAR = {2017}}

What makes a computational neuronal model `large scale'? Is it the number of neurons modeled? Or the number of brain regions modeled in a network? Most of the higher cognitive processes span across co-ordinated activity in a network of different brain areas. However at the same time, the basic information transfer takes place at a single neuron level, together with multiple other neurons. We explore modeling a neural system involving some areas of cortex, the basal ganglia (BG) and thalamus for the process of decision making, using a large-scale neural engineering framework, Nengo. Early results tend to replicate the known neural activity patterns as found in the previous action selection model by Guthrie et al. 2013, besides operating with a larger neuronal populations. The power of converting algorithms to efficiently weighed neural networks in Nengo (Stewart et al. 2009 and Bekolay et al. 2013) is …

@inproceedings{bib_A_bi_2017, AUTHOR = {KAUSHIK, PRAMOD SIVARAM and Carrere, Maxime and Alexandre, Fred´ Eric and Surampudi, Bapiraju }, TITLE = {A biologically inspired neuronal model of reward prediction error computation}, BOOKTITLE = {International Joint Conference on Neural Networks}. YEAR = {2017}}

The neurocomputational model described here proposes that two dimensions involved in computation of reward prediction errors i.e magnitude and time could be computed separately and later combined unlike traditional reinforcement learning models. The model is built on biological evidences and is able to reproduce various aspects of classical conditioning, namely, the progressive cancellation of the predicted reward, the predictive firing from conditioned stimuli, and delineation of early rewards by showing firing for sooner early rewards and not for early rewards that occur with a longer latency in accordance with biological data.

@inproceedings{bib_Lear_2017, AUTHOR = {MALU, GAUTAM KUMAR and Surampudi, Bapiraju and Indurkhya, Bipin }, TITLE = {Learning photography aesthetics with deep cnns}, BOOKTITLE = {Modern Artificial Intelligence and Cognitive Science Conference}. YEAR = {2017}}

utomatic photo aesthetic assessment is a challenging artificial intelligence task. Existing computational approaches have focused on modeling a single aesthetic score or a class (good or bad), however these do not provide any details on why the photograph is good or bad, or which attributes contribute to the quality of the photograph. To obtain both accuracy and human interpretation of the score, we advocate learning the aesthetic attributes along with the prediction of the overall score. For this purpose, we propose a novel multitask deep convolution neural network, which jointly learns eight aesthetic attributes along with the overall aesthetic score. We report near human performance in the prediction of the overall aesthetic score. To understand the internal representation of these attributes in the learned model, we also develop the visualization technique using back propagation of gradients. These visualizations highlight the important image regions for the corresponding attributes, thus providing insights about model's representation of these attributes. We showcase the diversity and complexity associated with different attributes through a qualitative analysis of the activation maps.

@inproceedings{bib_Iden_2017, AUTHOR = {SINGH, NOOR PRATAP and Surampudi, Bapiraju and Krishnanunni, Vinod Palakkad }, TITLE = {Identification of Biomarkers for Stage Prediction in Papillary Renal Cell Carcinoma}, BOOKTITLE = {Canadian Journal of Biotechnology}. YEAR = {2017}}

Papillary Renal Cell Carcinoma (PRCC) is a heterogeneous disease accounting for 10%-15% of renal cell carcinomas. A comprehensive analysis is required to find the genes that are responsible for the stage progression in PRCC. The advent of next generation sequencing techniques (NGS) has produced a lot of high throughput data from patients that can be analyzed to address this problem. The low sample size, noise and high dimensionality of the data though enhances the complexity, requiring the use of sophisticated methods. In our study we propose a machine learning pipeline fulfilling a two-fold objective: 1) To find suitable genes that could serve as potential biomarkers for stage progression in PRCC. 2) To build a classifier using the above biomarkers that can predict the stage of a given patient. The RNA-Seq data of PRCC was taken and divided into training set (80%) and test set (20%). Different …

@inproceedings{bib_Do_W_2017, AUTHOR = {Steephen, John Eric and MEHTA, RAJ and Surampudi, Bapiraju }, TITLE = {Do We Expect Women to Look Happier Than They Are? A Test of Gender-Dependent Perceptual Correction}, BOOKTITLE = {Perception}. YEAR = {2017}}

Feminine facial features enhance the expressive cues associated with happiness but not sadness. This makes a woman look happier than a man even when their smiles have the same intensity. So, to correctly infer the actual happiness of a woman, one would have to subtract the effect of these facial features. We hypothesised that our perceptual system would apply this subtraction for women, but not for men. This implies that this female-specific subtraction would cause one to infer a man to be happier than a woman if both are matched for facial appearance and expression intensity. We tested this using androgynous virtual faces with equal expression intensity. As predicted, happy men were inferred to be happier than happy women, but sad men were not inferred to be sadder than sad women, supporting our hypothesis of a gender- and emotion-specific perceptual correction.

@inproceedings{bib_Does_2016, AUTHOR = {Vattikonda, Anirudh and Surampudi, Bapiraju and Banerjee, Arpan and Deco, Gustavo and Roy, Dipanjan }, TITLE = {Does the regulation of local excitation–inhibition balance aid in recovery of functional connectivity? A computational account}, BOOKTITLE = {NeuroImage}. YEAR = {2016}}

Computational modeling of the spontaneous dynamics over the whole brain provides critical insight into the spatiotemporal organization of brain dynamics at multiple resolutions and their alteration to changes in brain structure (e.g. in diseased states, aging, across individuals). Recent experimental evidence further suggests that the adverse effect of lesions is visible on spontaneous dynamics characterized by changes in resting state functional connectivity and its graph theoretical properties (e.g. modularity). These changes originate from altered neural dynamics in individual brain areas that are otherwise poised towards a homeostatic equilibrium to maintain a stable excitatory and inhibitory activity. In this work, we employ a homeostatic inhibitory mechanism, balancing excitation and inhibition in the local brain areas of the entire cortex under neurological impairments like lesions to understand global functional …

@inproceedings{bib_Near_2016, AUTHOR = {DAGAR, SNIGDHA and Chowdhury, Shubhajit Roy and Surampudi, Bapiraju and Dutta, Anirban and Roy, Dipanjan }, TITLE = {Near-infrared spectroscopy–electroencephalography-based brain-state-dependent electrotherapy: a computational approach based on excitation–inhibition balance hypothesis}, BOOKTITLE = {Frontiers in Neurology}. YEAR = {2016}}

Stroke is the leading cause of severe chronic disability and the second cause of death worldwide with 15 million new cases and 50 million stroke survivors. The post stroke chronic disability may be ameliorated with early neuro rehabilitation where non-invasive brain stimulation (NIBS) techniques can be used as an adjuvant treatment to hasten the effects. However, the heterogeneity in the lesioned brain will require individualized NIBS intervention where innovative neuroimaging technologies of portable electroencephalography (EEG) and functional-near-infrared spectroscopy (fNIRS) can be leveraged for Brain State Dependent Electrotherapy (BSDE). In this hypothesis and theory article, we propose a computational approach based on excitation-inhibition (E-I) balance hypothesis to objectively quantify the post stroke individual brain state using online fNIRS-EEG joint imaging. One of the key events that occurs following Stroke is the imbalance in local excitation-inhibition (that is the ratio of Glutamate/GABA) which may be targeted with NIBS using a computational pipeline that includes individual “forward models” to predict current flow patterns through the lesioned brain or brain target region. The current flow will polarize the neurons which can be captured with excitation-inhibition based brain models. Furthermore, E-I balance hypothesis can be used to find the consequences of cellular polarization on neuronal information processing which can then be implicated in changes in function. We first review evidence that shows how this local imbalance between excitation-inhibition leading to functional

@inproceedings{bib_Inte_2016, AUTHOR = {Annam, Jagadeeswararao and Surampudi, Bapiraju }, TITLE = {Inter-patient heart-beat classification using complete ECG beat time series by alignment of R-peaks using SVM and decision rule}, BOOKTITLE = {International Conference on Signal and Information Processing}. YEAR = {2016}}

An ElectroCardiogram (ECG) inter-patient heartbeat time series classification method by a hierarchical system of based on support vector machine and Decision rule, using full heart-beat time series by alignment of R-peaks of all beats, is proposed. PQRST Time series of heart-beats having converted into equal length series by alignment of R-peaks of all heart-beats based on R-peak of largest length PQRST series in the data and by padding zeroes to the smaller length series on either side, was used in this experimentation. The main objective of this paper is to identify the abnormalities in ECG heart beats based on AAMI Categorization. Experiments were conducted on ECG data of 44 patients obtained from MIT-BIH Arrhythmia database. Results were compared with existing methods such as weighted support vector machine (SVM), hierarchical SVM and weighted linear discriminant analysis (LDA). Comparative …

@inproceedings{bib_A_un_2016, AUTHOR = {PARSOTTAM, SAVALIA TEJAS and SHUKLA, ANUJ KUMAR and Surampudi, Bapiraju }, TITLE = {A unified theoretical framework for cognitive sequencing}, BOOKTITLE = {Frontiers in psychology}. YEAR = {2016}}

The capacity to sequence information is central to human performance. Sequencing ability forms the foundation stone for higher order cognition related to language and goal-directed planning. Information related to the order of items, their timing, chunking and hierarchical organization are important aspects in sequencing. Past research on sequencing has emphasized two distinct and independent dichotomies: implicit versus explicit and goal-directed versus habits. We propose a theoretical framework unifying these two streams. Our proposal relies on brain's ability to implicitly extract statistical regularities from the stream of stimuli and with attentional engagement organizing sequences explicitly and hierarchically. Similarly, sequences that need to be assembled purposively to accomplish a goal require engagement of attentional processes. With repetition, these goal-directed plans become habits with concomitant disengagement of attention. Thus attention and awareness play a crucial role in the implicit-to-explicit transition as well as in how goal-directed plans become automatic habits. Cortico-subcortical loops ─ basal ganglia-frontal cortex and hippocampus-frontal cortex loops ─ mediate the transition process. We show how the computational principles of model-free and model-based learning paradigms, along with a pivotal role for attention and awareness, offer a unifying framework for these two dichotomies. Based on this framework, we make testable predictions related to the potential influence of response-to-stimulus interval (RSI) on developing awareness in implicit learning tasks.

@inproceedings{bib_AAMI_2016, AUTHOR = {, Jagadeeswara Rao Annam and Surampudi, Bapiraju }, TITLE = {AAMI based ECG heart-beat time-series clustering using unsupervised elm and decision rule}, BOOKTITLE = {International Conference on Information Technology}. YEAR = {2016}}

Early detection of cardiovascular diseases can prevent the premature deaths caused by abnormal heartbeat problems. Application of unsupervised classification by Extreme learning machine is addressed for ElectroCardiogram (ECG) heart-beat time series clustering by a hybrid of Extreme learning machine and Decision rule using full heart-beat time series by alignment of R-peaks of all beats is proposed in this work. PQRST Time series of heart-beats having converted into equal length series by alignment of R-peaks of all heart-beats based on R-peak of largest length PQRST series in the data and by padding zeroes to the smaller length series on either side, was used in this experimentation. The main objective of this paper is to identify the abnormalities in ECG heart beats based on AAMI Categorization. Because of the large patient specific characteristics in ECG heartbeat morphology across individuals, the …

@inproceedings{bib_Mode_2015, AUTHOR = {Krishnaiah, Sannapaneni and Surampudi, Bapiraju and Keeffe, Jill }, TITLE = {Modeling the risk of age-related macular degeneration and its predictive comparisons in a population in South India}, BOOKTITLE = {International Journal of Community Medicine and Public Health}. YEAR = {2015}}

Purpose: To model the modifiable risk factors by using the logistic regression (LR) and artificial neural network (ANN) models for prediction of progression of Age-related Macular Degeneration (AMD) and cross-validate these models for their predictive accuracies in a population in South India. Methods: The data (N= 3,723) were analyzed from Andhra Pradesh Eye Disease Study (APEDS) on participants aged≥ 40 years. Sub-population data from this sample were drawn by using Random under Sampling (RUS)(n= 213) and combination of RUS and Random over Sampling (ROS)(n= 1420) techniques. The modifiable and non-modifiable risk factors which were elicited as part of the study were used to derive the LR based risk score models and the model fit was assessed using bootstrap method for internal validity. The ANN model was built for three sets of data using the multi-layer feed-forward back propagation …

@inproceedings{bib_Arti_2015, AUTHOR = {Alshamiri, Abobakr Khalil and Singh, Alok and Surampudi, Bapiraju }, TITLE = {Artificial bee colony algorithm for clustering: an extreme learning approach}, BOOKTITLE = {Soft Computing}. YEAR = {2015}}

Extreme learning machine (ELM) as a new learning approach has shown its good generalization performance in regression and classification applications. Clustering analysis is an important tool to explore the structure of data and has been employed in many disciplines and applications. In this paper, we present a method that builds on ELM projection of input data into a high-dimensional feature space and followed by unsupervised clustering using artificial bee colony (ABC) algorithm. While ELM projection facilitates separability of clusters, a metaheuristic technique such as ABC algorithm overcomes problems of dependence on initialization of cluster centers and convergence to local minima suffered by conventional algorithms such as K-means. The proposed ELM-ABC algorithm is tested on 12 benchmark data sets. The experimental results show that the ELM-ABC algorithm can effectively improve the …

@inproceedings{bib_Expa_2015, AUTHOR = {Wutz, Andreas and SHUKLA, ANUJ KUMAR and Surampudi, Bapiraju and Melcher, David }, TITLE = {Expansion and Compression of Time Correlate with Information Processing in an Enumeration Task}, BOOKTITLE = {Plos One}. YEAR = {2015}}

Perception of temporal duration is subjective and is influenced by factors such as attention and context. For example, unexpected or emotional events are often experienced as if time subjectively expands, suggesting that the amount of information processed in a unit of time can be increased. Time dilation effects have been measured with an oddball paradigm in which an infrequent stimulus is perceived to last longer than standard stimuli in the rest of the sequence. Likewise, time compression for the oddball occurs when the duration of the standard items is relatively brief. Here, we investigated whether the amount of information processing changes when time is perceived as distorted. On each trial, an oddball stimulus of varying numerosity (1–14 items) and duration was presented along with standard items that were either short (70 ms) or long (1050 ms). Observers were instructed to count the number of dots within the oddball stimulus and to judge its relative duration with respect to the standards on that trial. Consistent with previous results, oddballs were reliably perceived as temporally distorted: expanded for longer standard stimuli blocks and compressed for shorter standards. The occurrence of these distortions of time perception correlated with perceptual processing; i.e. enumeration accuracy increased when time was perceived as expanded and decreased with temporal compression. These results suggest that subjective time distortions are not epiphenomenal, but reflect real changes in sensory processing. Such short-term plasticity in information processing rate could be evolutionarily advantageous in optimizing perception and action …

@inproceedings{bib_An_e_2015, AUTHOR = {Vemuri, Kavita and Surampudi, Bapiraju }, TITLE = {An exploratory investigation of functional network connectivity of empathy and default mode networks in a free-viewing task}, BOOKTITLE = {Brain Connectivity}. YEAR = {2015}}

This study reports dynamic functional network connectivity (dFNC) analysis on time courses of putative empathy networks—cognitive, emotional, and motor—and the default mode network (DMN) identified from independent components (ICs) derived by the group independent component analysis (ICA) method. The functional magnetic resonance imaging (fMRI) data were collected from 15 subjects watching movies of three genres, an animation (S1), Indian Hindi (S2), and a Hollywood English (S3) movie. The hypothesis of the study is that empathic engagement in a movie narrative would modulate the activation with the DMN. The clippings were individually rated for emotional expressions, context, and empathy self-response by the fMRI subjects post scanning and by 40 participants in an independent survey who rated at four time intervals in each clipping. The analysis illustrates the following: (a) the ICA method

@inproceedings{bib_Extr_2015, AUTHOR = {Vesapogu, Praveen Kumar and Surampudi, Bapiraju }, TITLE = {Extreme Learning Machine for Eukaryotic and Prokaryotic Promoter Prediction}, BOOKTITLE = {International Conference on Fuzzy and Neuro Computing}. YEAR = {2015}}

Promoters are DNA sequences containing regulatory elements required to guide and modulate the transcription initiation of the gene. Predicting promoter sequences in genomic sequences is a significant task in genome annotation and understanding transcriptional regulation. In the past decade many methods with many feature extraction schemes have been proposed for the prediction of eukaryotic and prokaryotic promoters. Still there is great need for more accurate and faster methods. In this paper we employed extreme learning machine algorithm (ELM), for promoter prediction in DNA sequences of H. sapiens, D. melanogaster, A. thaliana, C. elegans and E. coli. We extracted dinucleotide and CpG island features, and achieved accuracy above 90 % for all the five species. Performance is compared with the feed forward back propagation algorithm (BP) and support vector machines (SVM) and the …

@inproceedings{bib_Fact_2015, AUTHOR = {Nallapu, Bhargav Teja and Surampudi, Bapiraju and Rougier, Nicolas P }, TITLE = {Factors affecting reward based decision making: a computational study}, BOOKTITLE = {International Conference on Cognition, Brain and Computation}. YEAR = {2015}}

Task: The task is a probabilistic learning task as described in Pasquereau et al., 2007 where four visual cues are associated with different reward probabilities (0.00, 0.33, 0.66 & 1.00). A trial is made of the simultaneous presentation of two random cues with equal salience at two random positions. Some time after the presentation, a switch in the cortex activities is observed, representing the decision taken. After the model has chosen one cue or the other, a reward is given according to the probability associated with the chosen cue. Connections between the cortex and the striatum are then modified using a reinforcement learning rule based on the reward signal. The model is trained over 120 trials such that each combination of cues is presented equal number of times at uniformly sampled positions and the model performance reaches at least 0.9 measuring the ration of optimal choices. The decision switch and the performance are identical to the results when primates are tested with same task. Learning is then disabled and the model is tested using always the same pair of cues A (P (R)= 1) and B (P (R)= 0.33) in the presence of external factors. We study how, despite reward based learning, visual salience of the stimuli and the temporal difference between stimulus presentations affect the model to take a sub-optimal decision.

PhD Students (6)

MS Students (16)

Dual Degree Students (16)

Honours Students (6)