Signal detection theory, uncertainty, and Poisson-like population codes

Research output: Contribution to journalArticle

Abstract

The juxtaposition of established signal detection theory models of perception and more recent claims about the encoding of uncertainty in perception is a rich source of confusion. Are the latter simply a rehash of the former? Here, we make an attempt to distinguish precisely between optimal and probabilistic computation. In optimal computation, the observer minimizes the expected cost under a posterior probability distribution. In probabilistic computation, the observer uses higher moments of the likelihood function of the stimulus on a trial-by-trial basis. Computation can be optimal without being probabilistic, and vice versa. Most signal detection theory models describe optimal computation. Behavioral data only provide evidence for a neural representation of uncertainty if they are best described by a model of probabilistic computation. We argue that single-neuron activity sometimes suffices for optimal computation, but never for probabilistic computation. A population code is needed instead. Not every population code is equally suitable, because nuisance parameters have to be marginalized out. This problem is solved by Poisson-like, but not by Gaussian variability. Finally, we build a dictionary between signal detection theory quantities and Poisson-like population quantities.

Original languageEnglish (US)
Pages (from-to)2308-2319
Number of pages12
JournalVision Research
Volume50
Issue number22
DOIs
StatePublished - Oct 28 2010

Fingerprint

Uncertainty
Population
Likelihood Functions
Confusion
Statistical Models
Neurons
Costs and Cost Analysis
Psychological Signal Detection

Keywords

  • Bayesian inference
  • Population coding
  • Signal detection theory
  • Single neurons

ASJC Scopus subject areas

  • Ophthalmology
  • Sensory Systems

Cite this

Signal detection theory, uncertainty, and Poisson-like population codes. / Ma, Wei Ji.

In: Vision Research, Vol. 50, No. 22, 28.10.2010, p. 2308-2319.

Research output: Contribution to journalArticle

@article{2759ec50e48e416f8853e446d4cee149,
title = "Signal detection theory, uncertainty, and Poisson-like population codes",
abstract = "The juxtaposition of established signal detection theory models of perception and more recent claims about the encoding of uncertainty in perception is a rich source of confusion. Are the latter simply a rehash of the former? Here, we make an attempt to distinguish precisely between optimal and probabilistic computation. In optimal computation, the observer minimizes the expected cost under a posterior probability distribution. In probabilistic computation, the observer uses higher moments of the likelihood function of the stimulus on a trial-by-trial basis. Computation can be optimal without being probabilistic, and vice versa. Most signal detection theory models describe optimal computation. Behavioral data only provide evidence for a neural representation of uncertainty if they are best described by a model of probabilistic computation. We argue that single-neuron activity sometimes suffices for optimal computation, but never for probabilistic computation. A population code is needed instead. Not every population code is equally suitable, because nuisance parameters have to be marginalized out. This problem is solved by Poisson-like, but not by Gaussian variability. Finally, we build a dictionary between signal detection theory quantities and Poisson-like population quantities.",
keywords = "Bayesian inference, Population coding, Signal detection theory, Single neurons",
author = "Ma, {Wei Ji}",
year = "2010",
month = "10",
day = "28",
doi = "10.1016/j.visres.2010.08.035",
language = "English (US)",
volume = "50",
pages = "2308--2319",
journal = "Vision Research",
issn = "0042-6989",
publisher = "Elsevier Limited",
number = "22",

}

TY - JOUR

T1 - Signal detection theory, uncertainty, and Poisson-like population codes

AU - Ma, Wei Ji

PY - 2010/10/28

Y1 - 2010/10/28

N2 - The juxtaposition of established signal detection theory models of perception and more recent claims about the encoding of uncertainty in perception is a rich source of confusion. Are the latter simply a rehash of the former? Here, we make an attempt to distinguish precisely between optimal and probabilistic computation. In optimal computation, the observer minimizes the expected cost under a posterior probability distribution. In probabilistic computation, the observer uses higher moments of the likelihood function of the stimulus on a trial-by-trial basis. Computation can be optimal without being probabilistic, and vice versa. Most signal detection theory models describe optimal computation. Behavioral data only provide evidence for a neural representation of uncertainty if they are best described by a model of probabilistic computation. We argue that single-neuron activity sometimes suffices for optimal computation, but never for probabilistic computation. A population code is needed instead. Not every population code is equally suitable, because nuisance parameters have to be marginalized out. This problem is solved by Poisson-like, but not by Gaussian variability. Finally, we build a dictionary between signal detection theory quantities and Poisson-like population quantities.

AB - The juxtaposition of established signal detection theory models of perception and more recent claims about the encoding of uncertainty in perception is a rich source of confusion. Are the latter simply a rehash of the former? Here, we make an attempt to distinguish precisely between optimal and probabilistic computation. In optimal computation, the observer minimizes the expected cost under a posterior probability distribution. In probabilistic computation, the observer uses higher moments of the likelihood function of the stimulus on a trial-by-trial basis. Computation can be optimal without being probabilistic, and vice versa. Most signal detection theory models describe optimal computation. Behavioral data only provide evidence for a neural representation of uncertainty if they are best described by a model of probabilistic computation. We argue that single-neuron activity sometimes suffices for optimal computation, but never for probabilistic computation. A population code is needed instead. Not every population code is equally suitable, because nuisance parameters have to be marginalized out. This problem is solved by Poisson-like, but not by Gaussian variability. Finally, we build a dictionary between signal detection theory quantities and Poisson-like population quantities.

KW - Bayesian inference

KW - Population coding

KW - Signal detection theory

KW - Single neurons

UR - http://www.scopus.com/inward/record.url?scp=77957754433&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=77957754433&partnerID=8YFLogxK

U2 - 10.1016/j.visres.2010.08.035

DO - 10.1016/j.visres.2010.08.035

M3 - Article

VL - 50

SP - 2308

EP - 2319

JO - Vision Research

JF - Vision Research

SN - 0042-6989

IS - 22

ER -