Dissecting the roles of supervised and unsupervised learning in perceptual discrimination judgments

Yonatan Loewenstein; Ofri Raviv; Merav Ahissar

doi:10.1523/JNEUROSCI.0757-20.2020

Dissecting the roles of supervised and unsupervised learning in perceptual discrimination judgments

Yonatan Loewenstein^*, Ofri Raviv, Merav Ahissar

^*Corresponding author for this work

Research output: Contribution to journal › Article › peer-review

6 Scopus citations

Abstract

Our ability to compare sensory stimuli is a fundamental cognitive function, which is known to be affected by two biases: choice bias, which reflects a preference for a given response, and contraction bias, which reflects a tendency to perceive stimuli as similar to previous ones. To test whether both reflect supervised processes, we designed feedback protocols aimed to modify them and tested them in human participants. Choice bias was readily modifiable. However, contraction bias was not. To compare these results to those predicted from an optimal supervised process, we studied a noise-matched optimal linear discriminator (Perceptron). In this model, both biases were substantially modified, indicating that the "resilience"of contraction bias to feedback does not maximize performance. These results suggest that perceptual discrimination is a hierarchical, two-stage process. In the first, stimulus statistics are learned and integrated with representations in an unsupervised process that is impenetrable to external feedback. In the second, a binary judgment, learned in a supervised way, is applied to the combined percept.

Original language	American English
Pages (from-to)	757-765
Number of pages	9
Journal	Journal of Neuroscience
Volume	41
Issue number	4
DOIs	https://doi.org/10.1523/JNEUROSCI.0757-20.2020
State	Published - 27 Jan 2021

Bibliographical note

Funding Information:
Received Apr. 2, 2020; revised Oct. 26, 2020; accepted Nov. 5, 2020. Author contributions: Y.L., O.R., and M.A. designed research; O.R. performed research; Y.L., O.R., and M.A. analyzed data; Y.L., O.R., and M.A. wrote the paper. This work was supported by the European Research Council European Union’s Horizon 2020 Research and Innovation Program Grant 833694 (to M.A.) and the Israel Science Foundation Grants 757/16 (to Y.L.) and 1650/17 (to M.A.), the Deutsche Forschungsgemeinschaft Grant CRC 1080 (to Y.L.), and by the Gatsby Charitable Foundation (Y.L. and M.A.). The authors declare no competing financial interests. Correspondence should be addressed to Yonatan Loewenstein at yonatan@huji.ac.il. https://doi.org/10.1523/JNEUROSCI.0757-20.2020 Copyright © 2021 Loewenstein et al. This is an open-access article distributed under the terms of the Creative Commons Attribution 4.0 International license, which permits unrestricted use, distribution and reproduction in any medium provided that the original work is properly attributed.

Funding Information:
This work was supported by the European Research Council European Union's Horizon 2020 Research and Innovation Program Grant 833694 (to M.A.) and the Israel Science Foundation Grants 757/16 (to Y.L.) and 1650/17 (to M.A.), the Deutsche Forschungsgemeinschaft Grant CRC 1080 (to Y.L.), and by the Gatsby Charitable Foundation (Y.L. and M.A.).

Publisher Copyright:
Copyright © 2021 Loewenstein et al.

Keywords

Contraction bias
Frequency discrimination
Perception
Supervised learning
Unsupervised learning

Access to Document

10.1523/JNEUROSCI.0757-20.2020

Cite this

@article{4714ca26ab924e7d90926b580baf87e5,

title = "Dissecting the roles of supervised and unsupervised learning in perceptual discrimination judgments",

abstract = "Our ability to compare sensory stimuli is a fundamental cognitive function, which is known to be affected by two biases: choice bias, which reflects a preference for a given response, and contraction bias, which reflects a tendency to perceive stimuli as similar to previous ones. To test whether both reflect supervised processes, we designed feedback protocols aimed to modify them and tested them in human participants. Choice bias was readily modifiable. However, contraction bias was not. To compare these results to those predicted from an optimal supervised process, we studied a noise-matched optimal linear discriminator (Perceptron). In this model, both biases were substantially modified, indicating that the {"}resilience{"}of contraction bias to feedback does not maximize performance. These results suggest that perceptual discrimination is a hierarchical, two-stage process. In the first, stimulus statistics are learned and integrated with representations in an unsupervised process that is impenetrable to external feedback. In the second, a binary judgment, learned in a supervised way, is applied to the combined percept.",

keywords = "Contraction bias, Frequency discrimination, Perception, Supervised learning, Unsupervised learning",

author = "Yonatan Loewenstein and Ofri Raviv and Merav Ahissar",

note = "Funding Information: Received Apr. 2, 2020; revised Oct. 26, 2020; accepted Nov. 5, 2020. Author contributions: Y.L., O.R., and M.A. designed research; O.R. performed research; Y.L., O.R., and M.A. analyzed data; Y.L., O.R., and M.A. wrote the paper. This work was supported by the European Research Council European Union{\textquoteright}s Horizon 2020 Research and Innovation Program Grant 833694 (to M.A.) and the Israel Science Foundation Grants 757/16 (to Y.L.) and 1650/17 (to M.A.), the Deutsche Forschungsgemeinschaft Grant CRC 1080 (to Y.L.), and by the Gatsby Charitable Foundation (Y.L. and M.A.). The authors declare no competing financial interests. Correspondence should be addressed to Yonatan Loewenstein at yonatan@huji.ac.il. https://doi.org/10.1523/JNEUROSCI.0757-20.2020 Copyright {\textcopyright} 2021 Loewenstein et al. This is an open-access article distributed under the terms of the Creative Commons Attribution 4.0 International license, which permits unrestricted use, distribution and reproduction in any medium provided that the original work is properly attributed. Funding Information: This work was supported by the European Research Council European Union's Horizon 2020 Research and Innovation Program Grant 833694 (to M.A.) and the Israel Science Foundation Grants 757/16 (to Y.L.) and 1650/17 (to M.A.), the Deutsche Forschungsgemeinschaft Grant CRC 1080 (to Y.L.), and by the Gatsby Charitable Foundation (Y.L. and M.A.). Publisher Copyright: Copyright {\textcopyright} 2021 Loewenstein et al.",

year = "2021",

month = jan,

day = "27",

doi = "10.1523/JNEUROSCI.0757-20.2020",

language = "American English",

volume = "41",

pages = "757--765",

journal = "Journal of Neuroscience",

issn = "0270-6474",

publisher = "Society for Neuroscience",

number = "4",

}

TY - JOUR

T1 - Dissecting the roles of supervised and unsupervised learning in perceptual discrimination judgments

AU - Loewenstein, Yonatan

AU - Raviv, Ofri

AU - Ahissar, Merav

N1 - Funding Information: Received Apr. 2, 2020; revised Oct. 26, 2020; accepted Nov. 5, 2020. Author contributions: Y.L., O.R., and M.A. designed research; O.R. performed research; Y.L., O.R., and M.A. analyzed data; Y.L., O.R., and M.A. wrote the paper. This work was supported by the European Research Council European Union’s Horizon 2020 Research and Innovation Program Grant 833694 (to M.A.) and the Israel Science Foundation Grants 757/16 (to Y.L.) and 1650/17 (to M.A.), the Deutsche Forschungsgemeinschaft Grant CRC 1080 (to Y.L.), and by the Gatsby Charitable Foundation (Y.L. and M.A.). The authors declare no competing financial interests. Correspondence should be addressed to Yonatan Loewenstein at yonatan@huji.ac.il. https://doi.org/10.1523/JNEUROSCI.0757-20.2020 Copyright © 2021 Loewenstein et al. This is an open-access article distributed under the terms of the Creative Commons Attribution 4.0 International license, which permits unrestricted use, distribution and reproduction in any medium provided that the original work is properly attributed. Funding Information: This work was supported by the European Research Council European Union's Horizon 2020 Research and Innovation Program Grant 833694 (to M.A.) and the Israel Science Foundation Grants 757/16 (to Y.L.) and 1650/17 (to M.A.), the Deutsche Forschungsgemeinschaft Grant CRC 1080 (to Y.L.), and by the Gatsby Charitable Foundation (Y.L. and M.A.). Publisher Copyright: Copyright © 2021 Loewenstein et al.

PY - 2021/1/27

Y1 - 2021/1/27

N2 - Our ability to compare sensory stimuli is a fundamental cognitive function, which is known to be affected by two biases: choice bias, which reflects a preference for a given response, and contraction bias, which reflects a tendency to perceive stimuli as similar to previous ones. To test whether both reflect supervised processes, we designed feedback protocols aimed to modify them and tested them in human participants. Choice bias was readily modifiable. However, contraction bias was not. To compare these results to those predicted from an optimal supervised process, we studied a noise-matched optimal linear discriminator (Perceptron). In this model, both biases were substantially modified, indicating that the "resilience"of contraction bias to feedback does not maximize performance. These results suggest that perceptual discrimination is a hierarchical, two-stage process. In the first, stimulus statistics are learned and integrated with representations in an unsupervised process that is impenetrable to external feedback. In the second, a binary judgment, learned in a supervised way, is applied to the combined percept.

AB - Our ability to compare sensory stimuli is a fundamental cognitive function, which is known to be affected by two biases: choice bias, which reflects a preference for a given response, and contraction bias, which reflects a tendency to perceive stimuli as similar to previous ones. To test whether both reflect supervised processes, we designed feedback protocols aimed to modify them and tested them in human participants. Choice bias was readily modifiable. However, contraction bias was not. To compare these results to those predicted from an optimal supervised process, we studied a noise-matched optimal linear discriminator (Perceptron). In this model, both biases were substantially modified, indicating that the "resilience"of contraction bias to feedback does not maximize performance. These results suggest that perceptual discrimination is a hierarchical, two-stage process. In the first, stimulus statistics are learned and integrated with representations in an unsupervised process that is impenetrable to external feedback. In the second, a binary judgment, learned in a supervised way, is applied to the combined percept.

KW - Contraction bias

KW - Frequency discrimination

KW - Perception

KW - Supervised learning

KW - Unsupervised learning

UR - http://www.scopus.com/inward/record.url?scp=85101729024&partnerID=8YFLogxK

U2 - 10.1523/JNEUROSCI.0757-20.2020

DO - 10.1523/JNEUROSCI.0757-20.2020

M3 - Article

C2 - 33380471

AN - SCOPUS:85101729024

SN - 0270-6474

VL - 41

SP - 757

EP - 765

JO - Journal of Neuroscience

JF - Journal of Neuroscience

IS - 4

ER -

Dissecting the roles of supervised and unsupervised learning in perceptual discrimination judgments

Abstract

Bibliographical note

Keywords

Access to Document

Other files and links

Fingerprint

Cite this