In Theory IIb we characterize with a mix of theory and experiments the optimization of deep convolutional networks by Stochastic Gradient Descent. The main new result in this paper is theoretical and experimental evidence for the following conjecture about SGD: SGD concentrates in probability -- like the classical Langevin equation -- on large volume, "flat" minima, selecting flat minimizers which are with very high probability also global minimizers
@article{arxiv.1801.02254,
title = {Theory of Deep Learning IIb: Optimization Properties of SGD},
author = {Chiyuan Zhang and Qianli Liao and Alexander Rakhlin and Brando Miranda and Noah Golowich and Tomaso Poggio},
journal= {arXiv preprint arXiv:1801.02254},
year = {2018}
}