A note on overfitting properties of KIC and KICc

Abd Krim Seghouane

doi:10.1016/j.sigpro.2006.01.002

A note on overfitting properties of KIC and KIC_c

Abd Krim Seghouane^*

^*Corresponding author for this work

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

11 Citations (Scopus)

Abstract

The Kullback information criterion, KIC and its univariate bias-corrected version, KIC_c may be viewed as estimators of the expected Kullback-Leibler symmetric divergence. This correspondence examines the overfitting properties of KIC and KIC_c through the probabilities of overfitting both in finite samples and asymptotically. It is shown that KIC and KIC_c have much smaller probabilities of overfitting than the Akaike information criterion, AIC, and its bias-corrected version AIC_c.

Original language	English
Pages (from-to)	3055-3060
Number of pages	6
Journal	Signal Processing
Volume	86
Issue number	10
DOIs	https://doi.org/10.1016/j.sigpro.2006.01.002
Publication status	Published - Oct 2006

Access to Document

10.1016/j.sigpro.2006.01.002

Cite this

@article{cd2b81c9dc3c42e1883c6a455ba3cec4,

title = "A note on overfitting properties of KIC and KICc",

abstract = "The Kullback information criterion, KIC and its univariate bias-corrected version, KICc may be viewed as estimators of the expected Kullback-Leibler symmetric divergence. This correspondence examines the overfitting properties of KIC and KICc through the probabilities of overfitting both in finite samples and asymptotically. It is shown that KIC and KICc have much smaller probabilities of overfitting than the Akaike information criterion, AIC, and its bias-corrected version AICc.",

keywords = "AIC, KIC, Kullback-Leibler information, Model selection",

author = "Seghouane, {Abd Krim}",

year = "2006",

month = oct,

doi = "10.1016/j.sigpro.2006.01.002",

language = "English",

volume = "86",

pages = "3055--3060",

journal = "Signal Processing",

issn = "0165-1684",

publisher = "Elsevier B.V.",

number = "10",

}

TY - JOUR

T1 - A note on overfitting properties of KIC and KICc

AU - Seghouane, Abd Krim

PY - 2006/10

Y1 - 2006/10

N2 - The Kullback information criterion, KIC and its univariate bias-corrected version, KICc may be viewed as estimators of the expected Kullback-Leibler symmetric divergence. This correspondence examines the overfitting properties of KIC and KICc through the probabilities of overfitting both in finite samples and asymptotically. It is shown that KIC and KICc have much smaller probabilities of overfitting than the Akaike information criterion, AIC, and its bias-corrected version AICc.

AB - The Kullback information criterion, KIC and its univariate bias-corrected version, KICc may be viewed as estimators of the expected Kullback-Leibler symmetric divergence. This correspondence examines the overfitting properties of KIC and KICc through the probabilities of overfitting both in finite samples and asymptotically. It is shown that KIC and KICc have much smaller probabilities of overfitting than the Akaike information criterion, AIC, and its bias-corrected version AICc.

KW - AIC

KW - KIC

KW - Kullback-Leibler information

KW - Model selection

UR - http://www.scopus.com/inward/record.url?scp=33745852805&partnerID=8YFLogxK

U2 - 10.1016/j.sigpro.2006.01.002

DO - 10.1016/j.sigpro.2006.01.002

M3 - Article

SN - 0165-1684

VL - 86

SP - 3055

EP - 3060

JO - Signal Processing

JF - Signal Processing

IS - 10

ER -

A note on overfitting properties of KIC and KIC_c

Abstract

Access to Document

Other files and links

Fingerprint

Cite this