Current browse context:
cs.LG
Change to browse by:
References & Citations
Computer Science > Machine Learning
Title: Failures of Gradient-Based Deep Learning
(Submitted on 23 Mar 2017 (v1), last revised 26 Apr 2017 (this version, v2))
Abstract: In recent years, Deep Learning has become the go-to solution for a broad range of applications, often outperforming state-of-the-art. However, it is important, for both theoreticians and practitioners, to gain a deeper understanding of the difficulties and limitations associated with common approaches and algorithms. We describe four types of simple problems, for which the gradient-based algorithms commonly used in deep learning either fail or suffer from significant difficulties. We illustrate the failures through practical experiments, and provide theoretical insights explaining their source, and how they might be remedied.
Submission history
From: Shaked Shammah [view email][v1] Thu, 23 Mar 2017 07:16:37 GMT (316kb,D)
[v2] Wed, 26 Apr 2017 05:23:26 GMT (406kb,D)
Link back to: arXiv, form interface, contact.