Publications:
You can view my full list of publications at [Google Scholar] or [Semantic Scholar].
Thesis:
On the Sample Complexity of Reinforcement Learning
Sham M. Kakade
Gatsby Computational Neuroscience Unit
University College London, 2003
Book:
We are writing a monograph on Reinforcement Learning. We will be periodically making updates to the draft.
See here for the related course.
Blog Posts:
[How Does Critical Batch Size Scale in Pre-training?]
[Mixture of Parrots: Experts Improve Memorization More Than Reasoning]
[Anything but SGD: Evaluating Optimizers for LLM Training]
[Transcendence: Generative Models Can Outperform the Experts That Train Them]
[Repeat After Me: Transformers are Better than State Space Models at Copying]
[A Next-Generation Architecture for Elastic and Conditional Computation]