Recent & Upcoming Talks

As a part of the selection process for the Fellowship, I gave a 15 minutes talk about my current and future research. With many interesting discussions afterwards.

30 minutes oral talk on modern optimizers and their efficacy in LLM training.

30 minutes oral talk on Newton Method with rank-one approximations of the Hessian.

1 hour oral talk on the Defense against Feature Reconstruction attacks.

1.5 hour oral talk on Model and Data Reconstruction attacks in Horizontal Federated Learning.

1.5 hour oral talk on Concept Bottleneck Models and Contrastive Learning in explainable AI.