Relative deviation learning bounds and generalization with unbounded loss functions

Corinna Cortes; Spencer Greenberg; Mehryar Mohri

doi:10.1007/s10472-018-9613-y

Relative deviation learning bounds and generalization with unbounded loss functions

Corinna Cortes, Spencer Greenberg, Mehryar Mohri

Computer Science

Research output: Contribution to journal › Article › peer-review

Abstract

We present an extensive analysis of relative deviation bounds, including detailed proofs of two-sided inequalities and their implications. We also give detailed proofs of two-sided generalization bounds that hold in the general case of unbounded loss functions, under the assumption that a moment of the loss is bounded. We then illustrate how to apply these results in a sample application: the analysis of importance weighting.

Original language	English (US)
Pages (from-to)	45-70
Number of pages	26
Journal	Annals of Mathematics and Artificial Intelligence
Volume	85
Issue number	1
DOIs	https://doi.org/10.1007/s10472-018-9613-y
State	Published - Jan 1 2019

Keywords

Generalization bounds
Importance weighting
Learning theory
Machine learning
Relative deviation bounds
Unbounded loss functions
Unbounded regression

ASJC Scopus subject areas

Applied Mathematics
Artificial Intelligence

Access to Document

10.1007/s10472-018-9613-y

Cite this

@article{a82b4dd7aa174a81876fbfe3f496f18b,

title = "Relative deviation learning bounds and generalization with unbounded loss functions",

abstract = "We present an extensive analysis of relative deviation bounds, including detailed proofs of two-sided inequalities and their implications. We also give detailed proofs of two-sided generalization bounds that hold in the general case of unbounded loss functions, under the assumption that a moment of the loss is bounded. We then illustrate how to apply these results in a sample application: the analysis of importance weighting.",

keywords = "Generalization bounds, Importance weighting, Learning theory, Machine learning, Relative deviation bounds, Unbounded loss functions, Unbounded regression",

author = "Corinna Cortes and Spencer Greenberg and Mehryar Mohri",

note = "Publisher Copyright: {\textcopyright} 2019, Springer Nature Switzerland AG.",

year = "2019",

month = jan,

day = "1",

doi = "10.1007/s10472-018-9613-y",

language = "English (US)",

volume = "85",

pages = "45--70",

journal = "Annals of Mathematics and Artificial Intelligence",

issn = "1012-2443",

publisher = "Springer Netherlands",

number = "1",

}

TY - JOUR

T1 - Relative deviation learning bounds and generalization with unbounded loss functions

AU - Cortes, Corinna

AU - Greenberg, Spencer

AU - Mohri, Mehryar

PY - 2019/1/1

Y1 - 2019/1/1

N2 - We present an extensive analysis of relative deviation bounds, including detailed proofs of two-sided inequalities and their implications. We also give detailed proofs of two-sided generalization bounds that hold in the general case of unbounded loss functions, under the assumption that a moment of the loss is bounded. We then illustrate how to apply these results in a sample application: the analysis of importance weighting.

AB - We present an extensive analysis of relative deviation bounds, including detailed proofs of two-sided inequalities and their implications. We also give detailed proofs of two-sided generalization bounds that hold in the general case of unbounded loss functions, under the assumption that a moment of the loss is bounded. We then illustrate how to apply these results in a sample application: the analysis of importance weighting.

KW - Generalization bounds

KW - Importance weighting

KW - Learning theory

KW - Machine learning

KW - Relative deviation bounds

KW - Unbounded loss functions

KW - Unbounded regression

UR - http://www.scopus.com/inward/record.url?scp=85059762463&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85059762463&partnerID=8YFLogxK

U2 - 10.1007/s10472-018-9613-y

DO - 10.1007/s10472-018-9613-y

M3 - Article

AN - SCOPUS:85059762463

SN - 1012-2443

VL - 85

SP - 45

EP - 70

JO - Annals of Mathematics and Artificial Intelligence

JF - Annals of Mathematics and Artificial Intelligence

IS - 1

ER -

Relative deviation learning bounds and generalization with unbounded loss functions

Abstract

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this