Theory of Deep Learning IIb: Optimization Properties of SGD
Name
CBMM-Memo-072.pdf
Size
3.66 MB
Format
Adobe PDF
Checksum (MD5)
573db37df7dc4efde517d8e1ed45f370
Author(s)
Zhang, Chiyuan
Liao, Qianli
Rakhlin, Alexander
Miranda, Brando
Golowich, Noah
Poggio, Tomaso
Date Issued
December 27, 2017
Publisher
Center for Brains, Minds and Machines (CBMM)
Series/Report no.
CBMM Memo Series;072
Abstract
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.
Persistent DSpace Link
