Investigating properties and reward functions of GRPO for training small LLMs on a toy math
dataset. This is a Research Project at the University of Tübingen.
Status: Ongoing Research. Mentor:Dr. Antonio Orvieto,
Dr. Pavel Kolev Datasets: Simple Arithmetic dataset (Countdown: Perform arithmetic operations on given
list of numbers to obtain a pre-defined target.) Language/Frameworks: Python, PyTorch
Transformer and Graph-based Prediction of Mechnaism of Action in Drug-Target Interactions
Using multimodal data with graph neural networks and SOTA LLM-based molecule encoders for predicting
mechanism of action in drug-target interactions.
Transformer and Graph-based Prediction of Mechnaism of Action in Drug-Target Interactions
Description
Using multimodal data with graph neural networks for small molecules and SOTA LLM-based protein
and molecule encoders for predicting mechanism of action (MoA) in drug-target interactions. The
prediction of MoA is a novel work, never done before. We also use secondary structure
information of protein as a multimodal data source.
Status: Manuscript under review. Mentor:Rangan Das,
Prof. Dr. Ujjwal Maulik Datasets: Dataset developed with data from DrugBank, Uniprot and Alphafold. Other: KIBA,
DAVIS, HUMAN, BindingDB Language/Frameworks: Python, PyTorch
Sketch-based Multiview 3D Image Retrieval using Convolutions and Transformers
Using Convolution network + Vision Transformer for 2D sketch-based 3D image retrieval.
Sketch-based Multiview 3D Image Retrieval using Convolutions and Transformers
Description
Using Convolution network + Vision Transformer for 2D sketch-based 3D image retrieval.
Status: Currently archived. Work till successfully obtaining 2D camera angles of 3D
views and a basic model based on CLIP-pretrained ViT. Mentor:Dr. Biplab Banerjee Datasets: SHREC 2013, 2014 SBSR Language/Frameworks: Python, PyTorch
Generalization of Lottery Ticket Hypothesis + channel pruning technique for model compression
Using combined Knowledge Distillation + Lottery Ticket Hypothesis (LTH) + Channel Pruning for producing
faster computer vision models with lower number of parameters.
Generalization of Lottery Ticket Hypothesis + channel pruning technique for model compression
Description
Using combined Knowledge Distillation + Lottery Ticket Hypothesis (LTH) + Channel Pruning for
producing computer vision models with lower number of parameters.
Status: Successfully completed research at TCS Research and Innovation Labs Mentor:Dr. Nupur Sumeet Language/Frameworks: Python, PyTorch
Medical Image Segmentation in Radiology and Histopathology images
Developing fuzzy atrous convolutional layers for better image segmentation in medical datasets.
Study of novel anti-spoof detection methods using textures and other image processing
techniques.
Status: Currently archived. Completed with studies of contemporary and past image
processing ideas. Mentor:Prof. Sanjoy Kumar Saha Datasets: MSU-MFSD, CASIA-SURF CeFA Language/Frameworks: Python, PyTorch
Air pollution forecasting using spatio-temporal graph convolution networks
Time series based air pollution regression problem solved using spatio-temporal graph CNN.
Air pollution forecasting using spatio-temporal graph convolution networks
Description
Forecasting of air pollution concentrations across measuring stations (spatial) given a history
of pollution (temporal). Spatial layers use graph based structure simplified using Chebyshev's
polynomial approximation while temporal layers use 1D CNN and a GLU layer. Multi-pollutant
prediction simultaneously (multiple channels) is also explored.
Hand Gesture Recognition in Videos using OpenPose, 3D CNN and LSTM
Description
Developed during Google Summer of Code 2021 at Red Hen Lab. Detection of hand gestures using a
deep learning model consisting of LSTM and CNN. Hand keypoints obtained from
OpenPose.
ADDS - Attention-based Detection and Trajectory Prediction in Counter-Drone Systems
Description
Idea presentation at MLDS 2022 conference for using transformers in single drone trajectory
prediction system. Model design based on
Anticipative Video Transformer.
Paper:
link
Participatory Sensing Based Urban Road Condition Classification using Transfer Learning
Description
Image Classification algorithms such as VGG, ResNet, Inception, Xception, DenseNet, SEResNet,
and EfficientNet are applied on a road images dataset collected using participatory sensing.
Created the backend for the webapp using Node. Problems solved like basic CRUD operations with
MongoDB, providing shop-specified time slots to customers as per convenience and first-cum-first
basis. Changing of slots is also allowed.
Short introductory project on deep learning and tensorflow by implementing (with documentation)
of a simple art generator using neural style transfer using a style image and a source image.
