
Prasun Datta
PhD student · Rensselaer Polytechnic Institute
Works on the mathematical foundations of deep learning — how networks learn representations through gradient-based optimisation, why some learned features generalise, and how adversarial training reshapes hidden layers. The aim is rigorous theory for why deep networks work, when they fail, and how to make them reliable.
- Theoretical deep learning
- Optimization theory
- Representation learning
- Adversarial robustness






















