Please use this identifier to cite or link to this item: https://doi.org/10.1007/s10994-011-5268-1
Title: Robustness and generalization
Authors: Xu, H. 
Mannor, S.
Keywords: Generalization
Non-IID sample
Quantile loss
Robust
Issue Date: Mar-2012
Citation: Xu, H., Mannor, S. (2012-03). Robustness and generalization. Machine Learning 86 (3) : 391-423. ScholarBank@NUS Repository. https://doi.org/10.1007/s10994-011-5268-1
Abstract: We derive generalization bounds for learning algorithms based on their robustness: the property that if a testing sample is "similar" to a training sample, then the testing error is close to the training error. This provides a novel approach, different from complexity or stability arguments, to study generalization of learning algorithms. One advantage of the robustness approach, compared to previous methods, is the geometric intuition it conveys. Consequently, robustness-based analysis is easy to extend to learning in non-standard setups such as Markovian samples or quantile loss. We further show that a weak notion of robustness is both sufficient and necessary for generalizability, which implies that robustness is a fundamental property that is required for learning algorithms to work. © The Author(s) 2011.
Source Title: Machine Learning
URI: http://scholarbank.nus.edu.sg/handle/10635/85611
ISSN: 08856125
DOI: 10.1007/s10994-011-5268-1
Appears in Collections:Staff Publications

Show full item record
Files in This Item:
There are no files associated with this item.

SCOPUSTM   
Citations

45
checked on Dec 11, 2018

WEB OF SCIENCETM
Citations

29
checked on Dec 11, 2018

Page view(s)

36
checked on Dec 15, 2018

Google ScholarTM

Check

Altmetric


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.