Anmol Gupta

PhD Candidate at University of Groningen and IIT Roorkee.

About Me

Hi, this is Anmol. I am a research scholar doing joint PhD at University of Groningen, Netherlands and IIT Roorkee, India. I work under the supervision of Dr. Partha Pratim Roy, Dr. M.K. Van Vugt and Dr. Niels Taatgen. The topic of my PhD is “Tracking Depression and Elucidating its Mechanisms using Cognitive Neuroscience, EEG, and Machine Learning”. It is part of a project approved in the Scheme for Promotion of Academic and Collaboration (SPARC) by MHRD. My work includes Deep Learning, Computational Cognitive Neuroscience and Cognitive Modelling

Education

Joint PhD

University of Groningen and IIT Roorkee

2018 - 2022

Working on identifying and comparing different underlying mechanisms of depression using different modalities like behavioural and neurophysiological.

M. Tech.

National Institute of Technology, Hamirpur, India.

2014 - 2016

CGPA: 9.18, Worked on Automated Layout Generation of a Building using Genetic Algorithms in on order to assist an architect.

B. Tech.

National Institute of Technology, Uttarakhand, India.

2010 - 2014

CGPA: 8.93

XII | X

SJBSVM, Mathura, India

2009 | 2007

Marks: 89.4% in XII and 91.0% in X.

Publications

Evaluation of Instance-based Learning and Q-learning algorithms in dynamic environments

Click here to Read

A. Gupta, P.P. Roy and V. Dutt in IEEE Access

Reinforcement learning is an unsupervised learning algorithm, where learning is based upon feedback from the environment. Prior research has proposed cognitive (e.g., Instance-based Learning or IBL) and statistical (Q-learning) reinforcement learning algorithms. However, an evaluation of these algorithms in a single dynamic environment has not been explored. In this paper, a comparison between the statistical Q-learning algorithm and the cognitive IBL algorithm is presented. A well-known environment, “Frozen Lake,” is used to train, generalize, and scale Q-learning and IBL algorithms. For generalizing, the Q-learning and IBL agents were trained on one version of the Frozen Lake and tested on a permuted version of the same environment. For scaling, the two algorithms were tested on a larger version of the Frozen Lake environment. Results revealed that the IBL algorithm used less training time and generalized better to different environment variants. The IBL algorithm was also able to show scalability by retaining its superior performance in the larger environment. These results indicate that the IBL algorithm could be proposed as an alternative to the standard reinforcement learning algorithms based on dynamic programming such as Q-learning. The inclusion of human factors (such as memory) in the IBL algorithm makes it suitable for robust learning in complex and dynamic environments.

Subject-Specific Cognitive Workload Classification Using EEG-Based Functional Connectivity and Deep Learning

Click here to Read

A. Gupta, G. Siddhad, V. Pandey, P.P. Roy and B.G. Kim in MDPI Sensors

Cognitive workload is a crucial factor in tasks involving dynamic decision-making and other real-time and high-risk situations. Neuroimaging techniques have long been used for estimating cognitive workload. Given the portability, cost-effectiveness and high time-resolution of EEG as compared to fMRI and other neuroimaging modalities, an efficient method of estimating an individual’s workload using EEG is of paramount importance. Multiple cognitive, psychiatric and behavioral phenotypes have already been known to be linked with “functional connectivity”, i.e., correlations between different brain regions. In this work, we explored the possibility of using different model-free functional connectivity metrics along with deep learning in order to efficiently classify the cognitive workload of the participants. To this end, 64-channel EEG data of 19 participants were collected while they were doing the traditional n-back task. These data (after pre-processing) were used to extract the functional connectivity features, namely Phase Transfer Entropy (PTE), Mutual Information (MI) and Phase Locking Value (PLV). These three were chosen to do a comprehensive comparison of directed and non-directed model-free functional connectivity metrics (allows faster computations). Using these features, three deep learning classifiers, namely CNN, LSTM and Conv-LSTM were used for classifying the cognitive workload as low (1-back), medium (2-back) or high (3-back). With the high inter-subject variability in EEG and cognitive workload and recent research highlighting that EEG-based functional connectivity metrics are subject-specific, subject-specific classifiers were used. Results show the state-of-the-art multi-class classification accuracy with the combination of MI with CNN at 80.87%, followed by the combination of PLV with CNN (at 75.88%) and MI with LSTM (at 71.87%). The highest subject specific performance was achieved by the combinations of PLV with Conv-LSTM, and PLV with CNN with an accuracy of 97.92%, followed by the combination of MI with CNN (at 95.83%) and MI with Conv-LSTM (at 93.75%). The results highlight the efficacy of the combination of EEG-based model-free functional connectivity metrics and deep learning in order to classify cognitive workload. The work can further be extended to explore the possibility of classifying cognitive workload in real-time, dynamic and complex real-world scenarios.

Eeg-based age and gender prediction using deep blstm-lstm network model

Click here to Read

A. Gupta, P. Kaushik, PP. Roy and DP. Dogra in IEEE Sensor Letters

We classified age and gender of a person based on the baseline EEG Recordings. An industry standard EEG recording device has been used to record cerebral activities of 60 subjects (both male and female) in relaxed position with closed eyes. Deep BLSTM-LSTM network has been used to construct a hybrid learning framework for the classification. We achieved and accuracy of 93.7% and 97.5% for age and gender classification respectively. Our analysis also reveals that the beta band frequencies are better in predicting the age and gender as compared to other frequency bands of the EEG signals.

Internships and Visits

University of Groningen, Netherlands

Visiting PhD Candidate

March 21 - August 21

https://www.rug.nl/staff/m.k.van.vugt/

Comparing the underlying mechanisms of reward learning and spontaneous thoughts in detecting depression. A cross-national study in India and the USA.

We collected data of 61 participants from India and USA while they were doing the reward learning and spontaneous thinking tasks. We observed that reward learning is a better phenotype for dignosing depression and there are significant demographic factors that affect the prognosis of depression and closely related constructs like Rumination and Perseverative Thinking.

Bio-Medical Engineering Lab, INMAS, DRDO

Research Intern

Jul-Nov 2019

INMAS, DRDO, India.

Depression Detection with EEG data and studying the effects of meditation on Depression.

We collected EEG data of 27 Depressed patients before and after meditation (RajaYoga) from KGMC hospital, Karnal. The objective was to observe the before and after differences of meditation on depression and classifying it. We got promising results in differences between the connectivity matrices (DTF) before and after the meditation. The results indicate that decreased activity in the frontal areas of brain. Together with the decreaed questionnaire scores (Hamilton Depression Rating Scale (HDRS)) and the decreased frontal activity, it can be concluded that the mediation actually helped the depressed patients. We are still working on building the classifier.

ACS Lab, IIT Mandi

Research Intern

Feb-April 2019

ACS Lab

Compared reinforcement learning algorithms; Q-learning and Instance Based Learning.

We proposed cognitive models like PyIBL as an alternative approach to reinforcement learning. Did a comparison of PyIBL and Q-Learning on FrozenLake environment from OpenAI gym. The results suggest that PyIBL was able to outperform Q-learning in generalizing to a new environment and also took less training time in doing so. (Paper Submitted to IEEE Access)

Osaka Prefecture University, Japan.

Research Visit

May 2018

Osaka Prefecture University

Visited Osaka Prefecture University, Japan under the financial support of JST Sakura Science Plan to have technical discussions in the areas of brain computer interface and deep learning.

Projects

Interpreting Deep Learning Models

Specialization Certificate, Coursera.

Capstone Project for AI for Medicine Specialization on Coursera.

Used GradCam and Shaps for interpretation and feature importance of ResNet used for disease classification using stanford’s CheXpert X-ray dataset.

Search for Dark Matter hints with ML at CERN.

Specialization Certificate, Coursera.

Capstone Project for Advance Machine Learning Specialization on Coursera.

Used Bayesian optimization with Gaussian processes for SHiP detector design optimization to search for rare decay elements

Skills

Programming

Python, C++

Frameworks

Tensorflow, mne, gym

Tools

jupyter, git, docker

Volunteer Experience

Mentor and Alpha Tester for DeepLearning.ai at Coursera.

Voluntary Teaching

Certificate, Evidyalok.

Taught English to village kids at Sarath, Jharkhand using Evidyalok NGO Platform.

References

  • Dr. Varun Dutt, Associate Professor, IIT MANDI, varundutt@iitmandi.ac.in, 01905-267150
  • Dr. Kamlesh Dutta, Associate Professor, NIT HAMIRPUR, kd@nith.ac.in, 01972-254400/24