About

I am a PhD student based in Nice, France, affiliated with CNRS, 3IA, INRIA, COATI and Université Côte d’Azur. My supervisor is Prof. Emanuele Natale. My research focuses on the theoretical aspects of machine learning, specifically tackling problems at the intersection of theoretical computer science and physics. Drawing on my background in Physics (BS–MS, IISER Kolkata), my current interests center on foundational topics, including Hopfield networks and brain-inspired learning dynamics, the mathematical theory of neural network pruning, and the analysis of neural scaling laws.

Education

PhD, Theoretical Machine Learning
Inria
2025 — 2028 (expected)
BS-MS, Physics
Indian Institute of Science Education and Research, Kolkata
2020 — 2025

Research

Directed Hopfield Networks

I am working on the theory of Directed Hopfield networks (DHNs). DHNs are a variation of classic Hopfield networks where connections between neurons are not symmetric. This asymmetry prevents them from having a simple energy function, meaning they rarely settle into stable states. Instead, their dynamics can be much richer, producing cycles or even chaotic behavior. I am interested in the capacity of DHNs to store limit-cycle attractors and their robustness.

Strong Lottery Ticket Hypothesis

I have also been working on the Strong Lottery Ticket Hypothesis, which states that within a randomly initialized neural network, there exist subnetworks that can achieve performance comparable to a trained network without any training. Specifically, I have been interested in the interplay between weight quantization and pruning in the context of the Strong Lottery Ticket Hypothesis. See my Publications for more details.

Publications

Teaching

Contact