“MARINA: Faster Non-Convex Distributed Learning with Compression” has been accepted to ICML 2021.

Our paper “MARINA: Faster Non-Convex Distributed Learning with Compression” has been accepted to The Thirty-eighth International Conference on Machine Learning (ICML 2021) proceeding.

The arXiv link for the paper: https://arxiv.org/abs/2102.07845.

The acceptance rate from various news sources is described as 21.5%. It was my great pleasure to work with my peers: Eduard Gorbunov, Zhize Li, Peter Richtarik.

The YouTube recording of my short 5-minute talk is available here