Generalization in Deep Learning
Download all files submitted through automated deposit
Author(s)
Kawaguchi, Kenji
Kaelbling, Leslie Pack
Bengio, Yoshua
Advisor(s)
Leslie Kaelbling
Date Issued
May 1, 2018
Series/Report no.
MIT-CSAIL-TR-2018-014
Abstract
With a direct analysis of neural networks, this paper presents a mathematically tight generalization theory to partially address an open problem regarding the generalization of deep learning. Unlike previous bound-based theory, our main theory is quantitatively as tight as possible for every dataset individually, while producing qualitative insights competitively. Our results give insight into why and how deep learning can generalize well, despite its large capacity, complexity, possible algorithmic instability, nonrobustness, and sharp minima, answering to an open question in the literature. We also discuss limitations of our results and propose additional open problems.
Subjects
neural network
learning theory
Terms of Use
Creative Commons Attribution 4.0 International
Persistent DSpace Link
