IEEE - Institute of Electrical and Electronics Engineers, Inc. - Stochastic Gradient Descent Performs Variational Inference, Converges to Limit Cycles for Deep Networks

2018 Information Theory and Applications Workshop (ITA)

Author(s): Pratik Chaudhari ; Stefano Soatto
Publisher: IEEE - Institute of Electrical and Electronics Engineers, Inc.
Publication Date: 1 February 2018
Conference Location: San Diego, CA, USA
Conference Date: 11 February 2018
Page(s): 1 - 10
ISBN (Electronic): 978-1-7281-0124-8
DOI: 10.1109/ITA.2018.8503224
Regular:

Stochastic gradient descent (SGD) is widely believed to perform implicit regularization when used to train deep neural networks, but the precise manner in which this occurs has thus far been... View More

Advertisement