Our paper on anticorrelated noise injection got accepted to ICML 2022

AntiPGD moves to flat minima that generalize better

I’m happy that our paper on AntiPGD, a version of stochastic gradient descent using anticorrelated perturbations, got accpeted to ICML 2022 in Baltimore, USA. You can read the PDF here. Looking forward to attend my first ML conference after Covid, present this paper, and have some scientific discussions.

Hans Kersting
Hans Kersting
Postdoctoral Researcher

My research interests include Bayesian inference, dynamical systems, and optimization.