On Stability and Concentration of Measure

reportActive / Technical Report | Accession Number: ADA459829 | Open PDF

Abstract:

Stability conditions can be thought of as a way of controlling the variance of the learning process. Strong stability conditions additionally imply concentration of certain quantities around their expected values. It was shown recently that stability of learning algorithms is closely related to their generalization and consistency. In this paper we examine stability conditions from this point of view.

Author(s):
Rakhlin, Alexander ; Mukherjee, Sayan ; Poggio, Tomaso
Author Organization(s):
MASSACHUSETTS INST OF TECH CAMBRIDGE MA CENTER FOR BIOLOGICAL AND COMPUTATIONAL LEARNING
Descriptive Note:
Technical rept.
Supplementary Note:
Sponsored in part by the National Science Foundation Grants nos. IIS-0085836, IIS-0112991, IIS-0209289, EIA-0218693, EIA-0218506, and MH66239-01A1. DOI: 10.21236/ADA459829
Pagination:
0014

Security Markings

DOCUMENT & CONTEXTUAL SUMMARY

Distribution:
Approved For Public Release
Distribution Statement:
Approved For Public Release; Distribution Is Unlimited.

RECORD

Collection: TR
Identifying Numbers
Report Number(s):
CBCL-239
Contract/Grant Number(s):
N00014-00-1-0907, N00014-02-1-0915
Monitor Series:
ONR
 Subject Terms
Joint Capability Areas:
JCA_8_Building Partnerships
Modernization Areas:
AI and Machine Learning
Communities of Interest:
Biomedical
Descriptor(s):
*STABILITY, ALGORITHMS
Field(s)/Group(s):
Theoretical Mathematics
Keyword(s):
MCDIARMID S INEQUALITY, EMPRICAL RISK MINIMIZATION ALGORITHM, LEARNING ALGORITHMS
Report Date:
2004 Jun 01
Creation Date:
2007 Jan 24