Perturbation Models and PAC-Bayesian Generalization Bounds

Joseph Keshet; Subhransu Maji; Tamir Hazan; Tommi Jaakkola

doi:10.7551/mitpress/10761.003.0011

Perturbation Models and PAC-Bayesian Generalization Bounds

Joseph Keshet, Subhransu Maji, Tamir Hazan, Tommi Jaakkola

Technion - Israel Institute of Technology, Bar-Ilan University

Research output: Chapter in Book/Report/Conference proceeding › Chapter › peer-review

Abstract

In this chapter we explore the generalization power of perturbation models. Learning parameters that minimize the expected task loss of perturbation models amounts to minimizing PAC-Bayesian generalization bounds. We provide an elementary derivation of PAC-Bayesian generalization bounds, while focusing on their Bayesian components, namely their predictive probabilities and their posterior distributions.

Original language	English
Title of host publication	Perturbations, Optimization, and Statistics
Editors	T Hazan, G Papandreou, D Tarlow
Pages	289-309
Number of pages	21
DOIs	https://doi.org/10.7551/mitpress/10761.003.0011
State	Published - 1 Dec 2016

Access to Document

10.7551/mitpress/10761.003.0011

https://ieeexplore.ieee.org/abstract/document/8093864

Cite this

@inbook{f4ee01212f824ce08bbdaeb2dfbacc5b,

title = "Perturbation Models and PAC-Bayesian Generalization Bounds",

abstract = "In this chapter we explore the generalization power of perturbation models. Learning parameters that minimize the expected task loss of perturbation models amounts to minimizing PAC-Bayesian generalization bounds. We provide an elementary derivation of PAC-Bayesian generalization bounds, while focusing on their Bayesian components, namely their predictive probabilities and their posterior distributions.",

author = "Joseph Keshet and Subhransu Maji and Tamir Hazan and Tommi Jaakkola",

year = "2016",

month = dec,

day = "1",

doi = "10.7551/mitpress/10761.003.0011",

language = "الإنجليزيّة",

isbn = "9780262337939",

pages = "289--309",

editor = "T Hazan and G Papandreou and D Tarlow",

booktitle = "Perturbations, Optimization, and Statistics",

}

TY - CHAP

T1 - Perturbation Models and PAC-Bayesian Generalization Bounds

AU - Keshet, Joseph

AU - Maji, Subhransu

AU - Hazan, Tamir

AU - Jaakkola, Tommi

PY - 2016/12/1

Y1 - 2016/12/1

N2 - In this chapter we explore the generalization power of perturbation models. Learning parameters that minimize the expected task loss of perturbation models amounts to minimizing PAC-Bayesian generalization bounds. We provide an elementary derivation of PAC-Bayesian generalization bounds, while focusing on their Bayesian components, namely their predictive probabilities and their posterior distributions.

AB - In this chapter we explore the generalization power of perturbation models. Learning parameters that minimize the expected task loss of perturbation models amounts to minimizing PAC-Bayesian generalization bounds. We provide an elementary derivation of PAC-Bayesian generalization bounds, while focusing on their Bayesian components, namely their predictive probabilities and their posterior distributions.

UR - https://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=barilanwos&SrcAuth=WosAPI&KeyUT=WOS:000521530900011&DestLinkType=FullRecord&DestApp=WOS

U2 - 10.7551/mitpress/10761.003.0011

DO - 10.7551/mitpress/10761.003.0011

M3 - فصل

SN - 9780262337939

SP - 289

EP - 309

BT - Perturbations, Optimization, and Statistics

A2 - Hazan, T

A2 - Papandreou, G

A2 - Tarlow, D

ER -

Perturbation Models and PAC-Bayesian Generalization Bounds

Abstract

Access to Document

Other files and links

Fingerprint

Cite this