Conference blog.

I am attending Thirty-seventh Conference on Neural Information Processing Systems (Neurips2023) now and I am happy to share my experience with you all. I will write a short summary every day.

Theory Time

Two tutorials for today. The panel sessions after each tutorial are great and inspirational.

Talks I have attended

• Recent and Upcoming Developments in Randomized Numerical Linear Algebra for ML

TLDR: There is no popular random models yet in the field. The authors call for some user-friendly packages for researchers.

Comment: I like the brief explanation of traditional random LA (linear algebra) methods and the modern ones in machine learning. I also come over with similar problems where I need to design the experiments from scratches.

• Reconsidering Overfitting in the Age of Overparameterized Models (Highlight of my day)

TLDR: Understand harmless overfitting through the lens of over-parametrized linear regression model.

Comment: I have been working on this toopic and yet I stll find the talk interesting. The speaker managed a good tradeoff between high level explanation and math details.