A framework for testing identifiability of Bayesian models of perception

Luigi Acerbi, Wei Ji Ma, Sethu Vijayakumar

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Bayesian observer models are very effective in describing human performance in perceptual tasks, so much so that they are trusted to faithfully recover hidden mental representations of priors, likelihoods, or loss functions from the data. However, the intrinsic degeneracy of the Bayesian framework, as multiple combinations of elements can yield empirically indistinguishable results, prompts the question of model identifiability. We propose a novel framework for a systematic testing of the identifiability of a significant class of Bayesian observer models, with practical applications for improving experimental design. We examine the theoretical identifiability of the inferred internal representations in two case studies. First, we show which experimental designs work better to remove the underlying degeneracy in a time interval estimation task. Second, we find that the reconstructed representations in a speed perception task under a slow-speed prior are fairly robust.

Original languageEnglish (US)
Title of host publicationAdvances in Neural Information Processing Systems
PublisherNeural information processing systems foundation
Pages1026-1034
Number of pages9
Volume2
EditionJanuary
StatePublished - 2014
Event28th Annual Conference on Neural Information Processing Systems 2014, NIPS 2014 - Montreal, Canada
Duration: Dec 8 2014Dec 13 2014

Other

Other28th Annual Conference on Neural Information Processing Systems 2014, NIPS 2014
CountryCanada
CityMontreal
Period12/8/1412/13/14

Fingerprint

Design of experiments
Testing

ASJC Scopus subject areas

  • Computer Networks and Communications
  • Information Systems
  • Signal Processing

Cite this

Acerbi, L., Ma, W. J., & Vijayakumar, S. (2014). A framework for testing identifiability of Bayesian models of perception. In Advances in Neural Information Processing Systems (January ed., Vol. 2, pp. 1026-1034). Neural information processing systems foundation.

A framework for testing identifiability of Bayesian models of perception. / Acerbi, Luigi; Ma, Wei Ji; Vijayakumar, Sethu.

Advances in Neural Information Processing Systems. Vol. 2 January. ed. Neural information processing systems foundation, 2014. p. 1026-1034.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Acerbi, L, Ma, WJ & Vijayakumar, S 2014, A framework for testing identifiability of Bayesian models of perception. in Advances in Neural Information Processing Systems. January edn, vol. 2, Neural information processing systems foundation, pp. 1026-1034, 28th Annual Conference on Neural Information Processing Systems 2014, NIPS 2014, Montreal, Canada, 12/8/14.
Acerbi L, Ma WJ, Vijayakumar S. A framework for testing identifiability of Bayesian models of perception. In Advances in Neural Information Processing Systems. January ed. Vol. 2. Neural information processing systems foundation. 2014. p. 1026-1034
Acerbi, Luigi ; Ma, Wei Ji ; Vijayakumar, Sethu. / A framework for testing identifiability of Bayesian models of perception. Advances in Neural Information Processing Systems. Vol. 2 January. ed. Neural information processing systems foundation, 2014. pp. 1026-1034
@inproceedings{c7322284f92941c9b505c25bfd8c9025,
title = "A framework for testing identifiability of Bayesian models of perception",
abstract = "Bayesian observer models are very effective in describing human performance in perceptual tasks, so much so that they are trusted to faithfully recover hidden mental representations of priors, likelihoods, or loss functions from the data. However, the intrinsic degeneracy of the Bayesian framework, as multiple combinations of elements can yield empirically indistinguishable results, prompts the question of model identifiability. We propose a novel framework for a systematic testing of the identifiability of a significant class of Bayesian observer models, with practical applications for improving experimental design. We examine the theoretical identifiability of the inferred internal representations in two case studies. First, we show which experimental designs work better to remove the underlying degeneracy in a time interval estimation task. Second, we find that the reconstructed representations in a speed perception task under a slow-speed prior are fairly robust.",
author = "Luigi Acerbi and Ma, {Wei Ji} and Sethu Vijayakumar",
year = "2014",
language = "English (US)",
volume = "2",
pages = "1026--1034",
booktitle = "Advances in Neural Information Processing Systems",
publisher = "Neural information processing systems foundation",
edition = "January",

}

TY - GEN

T1 - A framework for testing identifiability of Bayesian models of perception

AU - Acerbi, Luigi

AU - Ma, Wei Ji

AU - Vijayakumar, Sethu

PY - 2014

Y1 - 2014

N2 - Bayesian observer models are very effective in describing human performance in perceptual tasks, so much so that they are trusted to faithfully recover hidden mental representations of priors, likelihoods, or loss functions from the data. However, the intrinsic degeneracy of the Bayesian framework, as multiple combinations of elements can yield empirically indistinguishable results, prompts the question of model identifiability. We propose a novel framework for a systematic testing of the identifiability of a significant class of Bayesian observer models, with practical applications for improving experimental design. We examine the theoretical identifiability of the inferred internal representations in two case studies. First, we show which experimental designs work better to remove the underlying degeneracy in a time interval estimation task. Second, we find that the reconstructed representations in a speed perception task under a slow-speed prior are fairly robust.

AB - Bayesian observer models are very effective in describing human performance in perceptual tasks, so much so that they are trusted to faithfully recover hidden mental representations of priors, likelihoods, or loss functions from the data. However, the intrinsic degeneracy of the Bayesian framework, as multiple combinations of elements can yield empirically indistinguishable results, prompts the question of model identifiability. We propose a novel framework for a systematic testing of the identifiability of a significant class of Bayesian observer models, with practical applications for improving experimental design. We examine the theoretical identifiability of the inferred internal representations in two case studies. First, we show which experimental designs work better to remove the underlying degeneracy in a time interval estimation task. Second, we find that the reconstructed representations in a speed perception task under a slow-speed prior are fairly robust.

UR - http://www.scopus.com/inward/record.url?scp=84937901706&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84937901706&partnerID=8YFLogxK

M3 - Conference contribution

VL - 2

SP - 1026

EP - 1034

BT - Advances in Neural Information Processing Systems

PB - Neural information processing systems foundation

ER -