DC Field | Value | Language |
---|---|---|
dc.contributor.author | Lee, Cheolhyung | ko |
dc.contributor.author | Cho, Kyunghyun | ko |
dc.contributor.author | Kang, Wanmo | ko |
dc.date.accessioned | 2019-01-23T05:15:56Z | - |
dc.date.available | 2019-01-23T05:15:56Z | - |
dc.date.created | 2018-12-13 | - |
dc.date.created | 2018-12-13 | - |
dc.date.created | 2018-12-13 | - |
dc.date.issued | 2018-12-08 | - |
dc.identifier.citation | Thirty-second Conference on Neural Information Processing Systems | - |
dc.identifier.uri | http://hdl.handle.net/10203/249201 | - |
dc.description.abstract | Although stochastic gradient descent (SGD) is a driving force behind the recent success of deep learning, our understanding of its dynamics in a high-dimensional parameter space is limited. In recent years, some researchers have used the stochasticity of minibatch gradients, or the signal-to-noise ratio, to better characterize the learning dynamics of SGD. Inspired from these work, we here analyze SGD from a geometrical perspective by inspecting the stochasticity of the norms and directions of minibatch gradients. We propose a model of the directional concentration for minibatch gradients through von Mises-Fisher (VMF) distribution, and show that the directional uniformity of minibatch gradients increases over the course of SGD. We empirically verify our result using deep convolutional networks and observe a higher correlation between the gradient stochasticity and the proposed directional uniformity than that against the gradient norm stochasticity, suggesting that the directional statistics of minibatch gradients is a major factor behind SGD. | - |
dc.language | English | - |
dc.publisher | Neural Information Processing Systems (NIPS) Foundation | - |
dc.title | Directional Analysis of Stochastic Gradient Descent via von Mises-Fisher Distributions in Deep Learning | - |
dc.type | Conference | - |
dc.type.rims | CONF | - |
dc.citation.publicationname | Thirty-second Conference on Neural Information Processing Systems | - |
dc.identifier.conferencecountry | CN | - |
dc.identifier.conferencelocation | Montreal Convention Center | - |
dc.contributor.localauthor | Kang, Wanmo | - |
dc.contributor.nonIdAuthor | Cho, Kyunghyun | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.