Pattern adaptation and normalization reweighting

Zachary M. Westrick, David Heeger, Michael Landy

Research output: Contribution to journalArticle

Abstract

Adaptation to an oriented stimulus changes both the gain and preferred orientation of neural responses in V1. Neurons tuned near the adapted orientation are suppressed, and their preferred orientations shift away from the adapter. We propose a model in which weights of divisive normalization are dynamically adjusted to homeostatically maintain response products between pairs of neurons. We demonstrate that this adjustment can be performed by a very simple learning rule. Simulations of this model closely match existing data from visual adaptation experiments. We consider several alternative models, including variants based on homeostatic maintenance of response correlations or covariance, as well as feedforward gain-control models with multiple layers, and we demonstrate that homeostatic maintenance of response products provides the best account of the physiological data.

Original languageEnglish (US)
Pages (from-to)9805-9816
Number of pages12
JournalJournal of Neuroscience
Volume36
Issue number38
DOIs
StatePublished - Sep 21 2016

Fingerprint

Maintenance
Social Adjustment
Neurons
Learning
Weights and Measures

Keywords

  • Adaptation
  • Hebbian learning
  • Normalization
  • V1

ASJC Scopus subject areas

  • Neuroscience(all)

Cite this

Pattern adaptation and normalization reweighting. / Westrick, Zachary M.; Heeger, David; Landy, Michael.

In: Journal of Neuroscience, Vol. 36, No. 38, 21.09.2016, p. 9805-9816.

Research output: Contribution to journalArticle

Westrick, Zachary M. ; Heeger, David ; Landy, Michael. / Pattern adaptation and normalization reweighting. In: Journal of Neuroscience. 2016 ; Vol. 36, No. 38. pp. 9805-9816.
@article{bb0421b8a4cf453caa9bcc77365f1f04,
title = "Pattern adaptation and normalization reweighting",
abstract = "Adaptation to an oriented stimulus changes both the gain and preferred orientation of neural responses in V1. Neurons tuned near the adapted orientation are suppressed, and their preferred orientations shift away from the adapter. We propose a model in which weights of divisive normalization are dynamically adjusted to homeostatically maintain response products between pairs of neurons. We demonstrate that this adjustment can be performed by a very simple learning rule. Simulations of this model closely match existing data from visual adaptation experiments. We consider several alternative models, including variants based on homeostatic maintenance of response correlations or covariance, as well as feedforward gain-control models with multiple layers, and we demonstrate that homeostatic maintenance of response products provides the best account of the physiological data.",
keywords = "Adaptation, Hebbian learning, Normalization, V1",
author = "Westrick, {Zachary M.} and David Heeger and Michael Landy",
year = "2016",
month = "9",
day = "21",
doi = "10.1523/JNEUROSCI.1067-16.2016",
language = "English (US)",
volume = "36",
pages = "9805--9816",
journal = "Journal of Neuroscience",
issn = "0270-6474",
publisher = "Society for Neuroscience",
number = "38",

}

TY - JOUR

T1 - Pattern adaptation and normalization reweighting

AU - Westrick, Zachary M.

AU - Heeger, David

AU - Landy, Michael

PY - 2016/9/21

Y1 - 2016/9/21

N2 - Adaptation to an oriented stimulus changes both the gain and preferred orientation of neural responses in V1. Neurons tuned near the adapted orientation are suppressed, and their preferred orientations shift away from the adapter. We propose a model in which weights of divisive normalization are dynamically adjusted to homeostatically maintain response products between pairs of neurons. We demonstrate that this adjustment can be performed by a very simple learning rule. Simulations of this model closely match existing data from visual adaptation experiments. We consider several alternative models, including variants based on homeostatic maintenance of response correlations or covariance, as well as feedforward gain-control models with multiple layers, and we demonstrate that homeostatic maintenance of response products provides the best account of the physiological data.

AB - Adaptation to an oriented stimulus changes both the gain and preferred orientation of neural responses in V1. Neurons tuned near the adapted orientation are suppressed, and their preferred orientations shift away from the adapter. We propose a model in which weights of divisive normalization are dynamically adjusted to homeostatically maintain response products between pairs of neurons. We demonstrate that this adjustment can be performed by a very simple learning rule. Simulations of this model closely match existing data from visual adaptation experiments. We consider several alternative models, including variants based on homeostatic maintenance of response correlations or covariance, as well as feedforward gain-control models with multiple layers, and we demonstrate that homeostatic maintenance of response products provides the best account of the physiological data.

KW - Adaptation

KW - Hebbian learning

KW - Normalization

KW - V1

UR - http://www.scopus.com/inward/record.url?scp=84988468570&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84988468570&partnerID=8YFLogxK

U2 - 10.1523/JNEUROSCI.1067-16.2016

DO - 10.1523/JNEUROSCI.1067-16.2016

M3 - Article

C2 - 27656020

AN - SCOPUS:84988468570

VL - 36

SP - 9805

EP - 9816

JO - Journal of Neuroscience

JF - Journal of Neuroscience

SN - 0270-6474

IS - 38

ER -