Decision-theoretic models of visual perception and action

Research output: Contribution to journalArticle

Abstract

Statistical decision theory (SDT) and Bayesian decision theory (BDT) are closely related mathematical frameworks used to model ideal performance in a wide range of visual and motor tasks. Their elements (gain function, likelihood, prior) are readily interpretable in terms of information available to the observer. We briefly describe SDT and BDT and then review recent work employing them as models of biological perception or action. We emphasize work that employs gain functions and priors as independent or dependent variables. At one extreme, Bayesian decision theory allows the experimenter to compute ideal performance in specific tasks and compare human performance to ideal (Geisler, 1989). No claim is made that visual processing is in any sense "Bayesian" At the other extreme, researchers have proposed Bayesian decision theory as a process model of "perception as Bayesian inference" (Knill & Richards, 1996). We end by discussing how possible ideal models are related to imperfect, actual observers and how the "Bayesian hypothesis" can be tested experimentally.

Original languageEnglish (US)
Pages (from-to)2362-2374
Number of pages13
JournalVision Research
Volume50
Issue number23
DOIs
StatePublished - Nov 23 2010

Fingerprint

Decision Theory
Visual Perception
Theoretical Models
Likelihood Functions
Biological Models
Research Personnel

Keywords

  • Action
  • Bayesian decision theory
  • Gain function
  • Ideal observer models
  • Likelihood
  • Loss function
  • Perception
  • Prior
  • Statistical decision theory

ASJC Scopus subject areas

  • Ophthalmology
  • Sensory Systems

Cite this

Decision-theoretic models of visual perception and action. / Maloney, Laurence T.; Zhang, Hang.

In: Vision Research, Vol. 50, No. 23, 23.11.2010, p. 2362-2374.

Research output: Contribution to journalArticle

Maloney, Laurence T. ; Zhang, Hang. / Decision-theoretic models of visual perception and action. In: Vision Research. 2010 ; Vol. 50, No. 23. pp. 2362-2374.
@article{cf48cc2fcf55441f8c458deaba4b299e,
title = "Decision-theoretic models of visual perception and action",
abstract = "Statistical decision theory (SDT) and Bayesian decision theory (BDT) are closely related mathematical frameworks used to model ideal performance in a wide range of visual and motor tasks. Their elements (gain function, likelihood, prior) are readily interpretable in terms of information available to the observer. We briefly describe SDT and BDT and then review recent work employing them as models of biological perception or action. We emphasize work that employs gain functions and priors as independent or dependent variables. At one extreme, Bayesian decision theory allows the experimenter to compute ideal performance in specific tasks and compare human performance to ideal (Geisler, 1989). No claim is made that visual processing is in any sense {"}Bayesian{"} At the other extreme, researchers have proposed Bayesian decision theory as a process model of {"}perception as Bayesian inference{"} (Knill & Richards, 1996). We end by discussing how possible ideal models are related to imperfect, actual observers and how the {"}Bayesian hypothesis{"} can be tested experimentally.",
keywords = "Action, Bayesian decision theory, Gain function, Ideal observer models, Likelihood, Loss function, Perception, Prior, Statistical decision theory",
author = "Maloney, {Laurence T.} and Hang Zhang",
year = "2010",
month = "11",
day = "23",
doi = "10.1016/j.visres.2010.09.031",
language = "English (US)",
volume = "50",
pages = "2362--2374",
journal = "Vision Research",
issn = "0042-6989",
publisher = "Elsevier Limited",
number = "23",

}

TY - JOUR

T1 - Decision-theoretic models of visual perception and action

AU - Maloney, Laurence T.

AU - Zhang, Hang

PY - 2010/11/23

Y1 - 2010/11/23

N2 - Statistical decision theory (SDT) and Bayesian decision theory (BDT) are closely related mathematical frameworks used to model ideal performance in a wide range of visual and motor tasks. Their elements (gain function, likelihood, prior) are readily interpretable in terms of information available to the observer. We briefly describe SDT and BDT and then review recent work employing them as models of biological perception or action. We emphasize work that employs gain functions and priors as independent or dependent variables. At one extreme, Bayesian decision theory allows the experimenter to compute ideal performance in specific tasks and compare human performance to ideal (Geisler, 1989). No claim is made that visual processing is in any sense "Bayesian" At the other extreme, researchers have proposed Bayesian decision theory as a process model of "perception as Bayesian inference" (Knill & Richards, 1996). We end by discussing how possible ideal models are related to imperfect, actual observers and how the "Bayesian hypothesis" can be tested experimentally.

AB - Statistical decision theory (SDT) and Bayesian decision theory (BDT) are closely related mathematical frameworks used to model ideal performance in a wide range of visual and motor tasks. Their elements (gain function, likelihood, prior) are readily interpretable in terms of information available to the observer. We briefly describe SDT and BDT and then review recent work employing them as models of biological perception or action. We emphasize work that employs gain functions and priors as independent or dependent variables. At one extreme, Bayesian decision theory allows the experimenter to compute ideal performance in specific tasks and compare human performance to ideal (Geisler, 1989). No claim is made that visual processing is in any sense "Bayesian" At the other extreme, researchers have proposed Bayesian decision theory as a process model of "perception as Bayesian inference" (Knill & Richards, 1996). We end by discussing how possible ideal models are related to imperfect, actual observers and how the "Bayesian hypothesis" can be tested experimentally.

KW - Action

KW - Bayesian decision theory

KW - Gain function

KW - Ideal observer models

KW - Likelihood

KW - Loss function

KW - Perception

KW - Prior

KW - Statistical decision theory

UR - http://www.scopus.com/inward/record.url?scp=78149402106&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=78149402106&partnerID=8YFLogxK

U2 - 10.1016/j.visres.2010.09.031

DO - 10.1016/j.visres.2010.09.031

M3 - Article

C2 - 20932856

AN - SCOPUS:78149402106

VL - 50

SP - 2362

EP - 2374

JO - Vision Research

JF - Vision Research

SN - 0042-6989

IS - 23

ER -