Majority-of-Three is Optimal [arXiv]
How simple can an optimal learner be?
TLDRs
Short, informal writeups of my papers, intended to require less background. I started writing these to get better at explaining things without relying as much on formalism; comments and suggestions are more than welcome :).
A full list of work is at research.
3️⃣
🐎
A Theory of Saddle Escape in Deep Nonlinear Networks [arXiv]
When do neural networks first learn features?
🔇
Rao-Blackwellized Score Matching on Manifolds [arXiv]
What do generative modeling techniques actually recover on structured data?