Estimating classification images with Generalized Linear and Additive Models

Kenneth Knoblauch, Laurence T. Maloney

Research output: Contribution to journalArticle

Abstract

Conventional approaches to modeling classification image data can be described in terms of a standard linear model (LM). We show how the problem can be characterized as a Generalized Linear Model (GLM) with a Bernoulli distribution. We demonstrate via simulation that this approach is more accurate in estimating the underlying template in the absence of internal noise. With increasing internal noise, however, the advantage of the GLM over the LM decreases and GLM is no more accurate than LM. We then introduce the Generalized Additive Model (GAM), an extension of GLM that can be used to estimate smooth classification images adaptively. We show that this approach is more robust to the presence of internal noise, and finally, we demonstrate that GAM is readily adapted to estimation of higher order (nonlinear) classification images and to testing their significance.

Original languageEnglish (US)
Article number10
JournalJournal of Vision
Volume8
Issue number16
DOIs
StatePublished - Dec 22 2008

Fingerprint

Linear Models
Noise
Binomial Distribution

Keywords

  • Classification images
  • GAM
  • Generalized additive models
  • Generalized linear models
  • GLM
  • Signal detection theory

ASJC Scopus subject areas

  • Ophthalmology
  • Sensory Systems

Cite this

Estimating classification images with Generalized Linear and Additive Models. / Knoblauch, Kenneth; Maloney, Laurence T.

In: Journal of Vision, Vol. 8, No. 16, 10, 22.12.2008.

Research output: Contribution to journalArticle

@article{e2613e22de3648c588182434d47f1480,
title = "Estimating classification images with Generalized Linear and Additive Models",
abstract = "Conventional approaches to modeling classification image data can be described in terms of a standard linear model (LM). We show how the problem can be characterized as a Generalized Linear Model (GLM) with a Bernoulli distribution. We demonstrate via simulation that this approach is more accurate in estimating the underlying template in the absence of internal noise. With increasing internal noise, however, the advantage of the GLM over the LM decreases and GLM is no more accurate than LM. We then introduce the Generalized Additive Model (GAM), an extension of GLM that can be used to estimate smooth classification images adaptively. We show that this approach is more robust to the presence of internal noise, and finally, we demonstrate that GAM is readily adapted to estimation of higher order (nonlinear) classification images and to testing their significance.",
keywords = "Classification images, GAM, Generalized additive models, Generalized linear models, GLM, Signal detection theory",
author = "Kenneth Knoblauch and Maloney, {Laurence T.}",
year = "2008",
month = "12",
day = "22",
doi = "10.1167/8.16.10",
language = "English (US)",
volume = "8",
journal = "Journal of Vision",
issn = "1534-7362",
publisher = "Association for Research in Vision and Ophthalmology Inc.",
number = "16",

}

TY - JOUR

T1 - Estimating classification images with Generalized Linear and Additive Models

AU - Knoblauch, Kenneth

AU - Maloney, Laurence T.

PY - 2008/12/22

Y1 - 2008/12/22

N2 - Conventional approaches to modeling classification image data can be described in terms of a standard linear model (LM). We show how the problem can be characterized as a Generalized Linear Model (GLM) with a Bernoulli distribution. We demonstrate via simulation that this approach is more accurate in estimating the underlying template in the absence of internal noise. With increasing internal noise, however, the advantage of the GLM over the LM decreases and GLM is no more accurate than LM. We then introduce the Generalized Additive Model (GAM), an extension of GLM that can be used to estimate smooth classification images adaptively. We show that this approach is more robust to the presence of internal noise, and finally, we demonstrate that GAM is readily adapted to estimation of higher order (nonlinear) classification images and to testing their significance.

AB - Conventional approaches to modeling classification image data can be described in terms of a standard linear model (LM). We show how the problem can be characterized as a Generalized Linear Model (GLM) with a Bernoulli distribution. We demonstrate via simulation that this approach is more accurate in estimating the underlying template in the absence of internal noise. With increasing internal noise, however, the advantage of the GLM over the LM decreases and GLM is no more accurate than LM. We then introduce the Generalized Additive Model (GAM), an extension of GLM that can be used to estimate smooth classification images adaptively. We show that this approach is more robust to the presence of internal noise, and finally, we demonstrate that GAM is readily adapted to estimation of higher order (nonlinear) classification images and to testing their significance.

KW - Classification images

KW - GAM

KW - Generalized additive models

KW - Generalized linear models

KW - GLM

KW - Signal detection theory

UR - http://www.scopus.com/inward/record.url?scp=58049087182&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=58049087182&partnerID=8YFLogxK

U2 - 10.1167/8.16.10

DO - 10.1167/8.16.10

M3 - Article

C2 - 19146276

AN - SCOPUS:58049087182

VL - 8

JO - Journal of Vision

JF - Journal of Vision

SN - 1534-7362

IS - 16

M1 - 10

ER -