Current browse context:
cs.LG
Change to browse by:
References & Citations
Computer Science > Machine Learning
Title: Adaptive Learning Rate via Covariance Matrix Based Preconditioning for Deep Neural Networks
(Submitted on 31 May 2016 (v1), last revised 28 Sep 2017 (this version, v2))
Abstract: Adaptive learning rate algorithms such as RMSProp are widely used for training deep neural networks. RMSProp offers efficient training since it uses first order gradients to approximate Hessian-based preconditioning. However, since the first order gradients include noise caused by stochastic optimization, the approximation may be inaccurate. In this paper, we propose a novel adaptive learning rate algorithm called SDProp. Its key idea is effective handling of the noise by preconditioning based on covariance matrix. For various neural networks, our approach is more efficient and effective than RMSProp and its variant.
Submission history
From: Yasutoshi Ida [view email][v1] Tue, 31 May 2016 12:11:51 GMT (29kb)
[v2] Thu, 28 Sep 2017 10:24:50 GMT (48kb,D)
Link back to: arXiv, form interface, contact.