Bayesian learning of visual chunks by human observers. Orbán, G., Fiser, J., Aslin, R. N, & Lengyel, M. Proc Natl Acad Sci U S A, 105(7):2745–2750, 2008. doi abstract bibtex 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.
@Article{Orban2008,
author = {Gergo Orb\'an and J\'ozsef Fiser and Richard N Aslin and M\'at\'e Lengyel},
journal = {Proc Natl Acad Sci U S A},
title = {Bayesian learning of visual chunks by human observers.},
year = {2008},
number = {7},
pages = {2745--2750},
volume = {105},
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.},
doi = {10.1073/pnas.0708424105},
institution = {Collegium Budapest Institute for Advanced Study, 2 Szenth\'aroms\'ag utca, Budapest H-1014, Hungary.},
keywords = {Bayes Theorem; Humans; Learning, physiology; Vision, Ocular, physiology},
language = {eng},
medline-pst = {ppublish},
pmid = {18268353},
timestamp = {2011.05.10},
}
Downloads: 0
{"_id":"JsvDpDk4Chou6Mjdv","bibbaseid":"orbn-fiser-aslin-lengyel-bayesianlearningofvisualchunksbyhumanobservers-2008","author_short":["Orbán, G.","Fiser, J.","Aslin, R. N","Lengyel, M."],"bibdata":{"bibtype":"article","type":"article","author":[{"firstnames":["Gergo"],"propositions":[],"lastnames":["Orbán"],"suffixes":[]},{"firstnames":["József"],"propositions":[],"lastnames":["Fiser"],"suffixes":[]},{"firstnames":["Richard","N"],"propositions":[],"lastnames":["Aslin"],"suffixes":[]},{"firstnames":["Máté"],"propositions":[],"lastnames":["Lengyel"],"suffixes":[]}],"journal":"Proc Natl Acad Sci U S A","title":"Bayesian learning of visual chunks by human observers.","year":"2008","number":"7","pages":"2745–2750","volume":"105","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.","doi":"10.1073/pnas.0708424105","institution":"Collegium Budapest Institute for Advanced Study, 2 Szentháromság utca, Budapest H-1014, Hungary.","keywords":"Bayes Theorem; Humans; Learning, physiology; Vision, Ocular, physiology","language":"eng","medline-pst":"ppublish","pmid":"18268353","timestamp":"2011.05.10","bibtex":"@Article{Orban2008,\n author = {Gergo Orb\\'an and J\\'ozsef Fiser and Richard N Aslin and M\\'at\\'e Lengyel},\n journal = {Proc Natl Acad Sci U S A},\n title = {Bayesian learning of visual chunks by human observers.},\n year = {2008},\n number = {7},\n pages = {2745--2750},\n volume = {105},\n abstract = {Efficient and versatile processing of any hierarchically structured\n\tinformation requires a learning mechanism that combines lower-level\n\tfeatures into higher-level chunks. We investigated this chunking\n\tmechanism in humans with a visual pattern-learning paradigm. We developed\n\tan ideal learner based on Bayesian model comparison that extracts\n\tand stores only those chunks of information that are minimally sufficient\n\tto encode a set of visual scenes. Our ideal Bayesian chunk learner\n\tnot only reproduced the results of a large set of previous empirical\n\tfindings in the domain of human pattern learning but also made a\n\tkey prediction that we confirmed experimentally. In accordance with\n\tBayesian learning but contrary to associative learning, human performance\n\twas well above chance when pair-wise statistics in the exemplars\n\tcontained no relevant information. Thus, humans extract chunks from\n\tcomplex visual patterns by generating accurate yet economical representations\n\tand not by encoding the full correlational structure of the input.},\n doi = {10.1073/pnas.0708424105},\n institution = {Collegium Budapest Institute for Advanced Study, 2 Szenth\\'aroms\\'ag utca, Budapest H-1014, Hungary.},\n keywords = {Bayes Theorem; Humans; Learning, physiology; Vision, Ocular, physiology},\n language = {eng},\n medline-pst = {ppublish},\n pmid = {18268353},\n timestamp = {2011.05.10},\n}\n\n","author_short":["Orbán, G.","Fiser, J.","Aslin, R. N","Lengyel, M."],"key":"Orban2008","id":"Orban2008","bibbaseid":"orbn-fiser-aslin-lengyel-bayesianlearningofvisualchunksbyhumanobservers-2008","role":"author","urls":{},"keyword":["Bayes Theorem; Humans; Learning","physiology; Vision","Ocular","physiology"],"metadata":{"authorlinks":{}}},"bibtype":"article","biburl":"http://endress.org/publications/ansgar.bib","dataSources":["xPGxHAeh3vZpx4yyE"],"keywords":["bayes theorem; humans; learning","physiology; vision","ocular","physiology"],"search_terms":["bayesian","learning","visual","chunks","human","observers","orbán","fiser","aslin","lengyel"],"title":"Bayesian learning of visual chunks by human observers.","year":2008}