Martin Marek

I am a PhD student in Machine Learning at NYU, working with Andrew Gordon Wilson and Pavel Izmailov. My research focuses on empirical contributions to the science of deep learning.

Previously I studied Statistics and Machine Learning at UCL and interned as a Quant Researcher at QRT and ML Engineer at Snapchat.

Links

GitHub, Google Scholar, Twitter

Email me: martin.m@nyu.edu

Personal projects

Burst Photo
📸 Mac app that brings “night mode” to any camera.
Implements HDR+ from scratch in Metal Shading Language.
[Website, GitHub, DPReview, PetaPixel, SonyAlphaRumors]

Selected papers

Small Batch Size Training for Language Models:
When Vanilla SGD Works, and Why Gradient Accumulation Is Wasteful
M. Marek, S. Lotfi, A. Somasundaram, A. G. Wilson, M. Goldblum
[PDF, arXiv, GitHub, Twitter, Blog]
Can a Confident Prior Replace a Cold Posterior?
M. Marek, B. Paige, P. Izmailov
[PDF, arXiv, GitHub, Twitter]

Website design inspired by Pavel Izmailov.