Dr. Leena Chennuru Vankadara

Lecturer at the Gatsby Computational Neuroscience Unit at UCL

Speaker: Dr. Leena Chennuru Vankadara Date: 11-02-2026, 11am-12pm (BST) Location: Department of Computer Science CS1.04, University of Warwick, Coventry, UK

Towards a theory of scaling in deep learning

Abstract

Scaling computational resources is essential for advancing modern deep learning, enabling both quantitative improvements and qualitative emergence of new behaviors. This talk introduces recent theoretical progress on using scaling limits to derive principled scaling rules demonstrating their impact on stability, feature learning, and hyperparameter predictability. We further address discrepancies between infinite-width theory and practice under standard parameterization (SP), revealing a novel “controlled divergence” regime that explains empirical phenomena including stable feature evolution and learning rate scaling behavior.

---

About Dr. Leena Chennuru Vankadara

Leena Chennuru Vankadara is a Lecturer at the Gatsby Computational Neuroscience Unit at UCL. Previously, she was an Applied Scientist at Amazon Research. She earned her Ph.D. in Theory of Machine Learning from the International Max Planck Institute for Intelligent Systems. Her research focuses on the theory of deep learning, including scaling, generalization, and learning dynamics.