Bayesian learning of visual chunks by human observers

Gergö Orbán, J. Fiser, Richard N. Aslin, Máté Lengyel

Research output: Contribution to journalArticle

114 Citations (Scopus)

Abstract

Efficient and versatile processing of any hierarchically structured information requires a learning mechanism that combines lower-level features into higher-level chunks. We investigated this chunking mechanism in humans with a visual pattern-learning paradigm. We developed an ideal learner based on Bayesian model comparison that extracts and stores only those chunks of information that are minimally sufficient to encode a set of visual scenes. Our ideal Bayesian chunk learner not only reproduced the results of a large set of previous empirical findings in the domain of human pattern learning but also made a key prediction that we confirmed experimentally. In accordance with Bayesian learning but contrary to associative learning, human performance was well above chance when pair-wise statistics in the exemplars contained no relevant information. Thus, humans extract chunks from complex visual patterns by generating accurate yet economical representations and not by encoding the full correlational structure of the input.

Original languageEnglish
Pages (from-to)2745-2750
Number of pages6
JournalProceedings of the National Academy of Sciences of the United States of America
Volume105
Issue number7
DOIs
Publication statusPublished - Feb 19 2008

Fingerprint

Learning
Complex Mixtures

Keywords

  • Bayesian inference
  • Probabilistic modeling
  • Vision

ASJC Scopus subject areas

  • Genetics
  • General

Cite this

Bayesian learning of visual chunks by human observers. / Orbán, Gergö; Fiser, J.; Aslin, Richard N.; Lengyel, Máté.

In: Proceedings of the National Academy of Sciences of the United States of America, Vol. 105, No. 7, 19.02.2008, p. 2745-2750.

Research output: Contribution to journalArticle

Orbán, Gergö ; Fiser, J. ; Aslin, Richard N. ; Lengyel, Máté. / Bayesian learning of visual chunks by human observers. In: Proceedings of the National Academy of Sciences of the United States of America. 2008 ; Vol. 105, No. 7. pp. 2745-2750.
@article{718b9bbbef3546819d7dab3d8cf5a68f,
title = "Bayesian learning of visual chunks by human observers",
abstract = "Efficient and versatile processing of any hierarchically structured information requires a learning mechanism that combines lower-level features into higher-level chunks. We investigated this chunking mechanism in humans with a visual pattern-learning paradigm. We developed an ideal learner based on Bayesian model comparison that extracts and stores only those chunks of information that are minimally sufficient to encode a set of visual scenes. Our ideal Bayesian chunk learner not only reproduced the results of a large set of previous empirical findings in the domain of human pattern learning but also made a key prediction that we confirmed experimentally. In accordance with Bayesian learning but contrary to associative learning, human performance was well above chance when pair-wise statistics in the exemplars contained no relevant information. Thus, humans extract chunks from complex visual patterns by generating accurate yet economical representations and not by encoding the full correlational structure of the input.",
keywords = "Bayesian inference, Probabilistic modeling, Vision",
author = "Gerg{\"o} Orb{\'a}n and J. Fiser and Aslin, {Richard N.} and M{\'a}t{\'e} Lengyel",
year = "2008",
month = "2",
day = "19",
doi = "10.1073/pnas.0708424105",
language = "English",
volume = "105",
pages = "2745--2750",
journal = "Proceedings of the National Academy of Sciences of the United States of America",
issn = "0027-8424",
number = "7",

}

TY - JOUR

T1 - Bayesian learning of visual chunks by human observers

AU - Orbán, Gergö

AU - Fiser, J.

AU - Aslin, Richard N.

AU - Lengyel, Máté

PY - 2008/2/19

Y1 - 2008/2/19

N2 - Efficient and versatile processing of any hierarchically structured information requires a learning mechanism that combines lower-level features into higher-level chunks. We investigated this chunking mechanism in humans with a visual pattern-learning paradigm. We developed an ideal learner based on Bayesian model comparison that extracts and stores only those chunks of information that are minimally sufficient to encode a set of visual scenes. Our ideal Bayesian chunk learner not only reproduced the results of a large set of previous empirical findings in the domain of human pattern learning but also made a key prediction that we confirmed experimentally. In accordance with Bayesian learning but contrary to associative learning, human performance was well above chance when pair-wise statistics in the exemplars contained no relevant information. Thus, humans extract chunks from complex visual patterns by generating accurate yet economical representations and not by encoding the full correlational structure of the input.

AB - Efficient and versatile processing of any hierarchically structured information requires a learning mechanism that combines lower-level features into higher-level chunks. We investigated this chunking mechanism in humans with a visual pattern-learning paradigm. We developed an ideal learner based on Bayesian model comparison that extracts and stores only those chunks of information that are minimally sufficient to encode a set of visual scenes. Our ideal Bayesian chunk learner not only reproduced the results of a large set of previous empirical findings in the domain of human pattern learning but also made a key prediction that we confirmed experimentally. In accordance with Bayesian learning but contrary to associative learning, human performance was well above chance when pair-wise statistics in the exemplars contained no relevant information. Thus, humans extract chunks from complex visual patterns by generating accurate yet economical representations and not by encoding the full correlational structure of the input.

KW - Bayesian inference

KW - Probabilistic modeling

KW - Vision

UR - http://www.scopus.com/inward/record.url?scp=40649090888&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=40649090888&partnerID=8YFLogxK

U2 - 10.1073/pnas.0708424105

DO - 10.1073/pnas.0708424105

M3 - Article

VL - 105

SP - 2745

EP - 2750

JO - Proceedings of the National Academy of Sciences of the United States of America

JF - Proceedings of the National Academy of Sciences of the United States of America

SN - 0027-8424

IS - 7

ER -