Designing and understanding fundamental limits of online privacy attack

Present (Under supervision of Prof. Debabrota Basu (INRIA, France))

Understanding and finding upper and lower bounds of privacy and adversarial attack advantage for online learning systems

Design of generalised attack set ups to encapsulate Membership Attack, Attribute Attack and Reconstruction attack

Application of Barlow Twins to train an encoder for NLP classification tasks using Self Supervised Learning

Present

Provide with a framework to train a robust encoder in self-supervised way to encode text for classification tasks

Text Generation From Knowledge Graph using Graph Transformers (Reproduction of a work by Rik et al.)

Mar 2021-June 2021

Leveraging Knowledge Graphs obtained from abstracts and titles of Research Papers in Computer Science (dataset called AGENDA) to train a Transformer based model that can generate Abstracts once the title and the knowledge graph for the abstract is given as input

Optimisation Algorithms Insights

Aug 2021

Analysis of optimization algorithms used in Machine Learning and Deep learning through visualisation (in Python)

A lucid visual walk through Fourier Transform

Apr 2021

Analysis and explanation of Fourier Transform through visualization in Python