Neural quadratic discriminant analysis: Nonlinear decoding with V1-like computation

Marino Pagan, Eero Simoncelli, Nicole C. Rust

Research output: Contribution to journalArticle

Abstract

Linear-nonlinear (LN) models and their extensions have proven successful in describing transformations from stimuli to spiking responses of neurons in early stages of sensory hierarchies. Neural responses at later stages are highly nonlinear and have generally been better characterized in terms of their decoding performance on prespecified tasks. Here we develop a biologically plausible decoding model for classification tasks, that we refer to as neural quadratic discriminant analysis (nQDA). Specifically, we reformulate an optimal quadratic classifier as an LN-LN computation, analogous to "subunit" encoding models that have been used to describe responses in retina and primary visual cortex. We propose a physiological mechanism by which the parameters of the nQDA classifier could be optimized, using a supervised variant of a Hebbian learning rule. As an example of its applicability, we show that nQDA provides a better account than many comparable alternatives for the transformation between neural representations in two high-level brain areas recorded as monkeys performed a visual delayed-match-to-sample task.

Original languageEnglish (US)
Pages (from-to)2291-2319
Number of pages29
JournalNeural Computation
Volume28
Issue number11
DOIs
StatePublished - Nov 1 2016

Fingerprint

Discriminant Analysis
Nonlinear Dynamics
Visual Cortex
Haplorhini
Retina
Linear Models
Learning
Neurons
Brain
Decoding
Classifier

ASJC Scopus subject areas

  • Arts and Humanities (miscellaneous)
  • Cognitive Neuroscience

Cite this

Neural quadratic discriminant analysis : Nonlinear decoding with V1-like computation. / Pagan, Marino; Simoncelli, Eero; Rust, Nicole C.

In: Neural Computation, Vol. 28, No. 11, 01.11.2016, p. 2291-2319.

Research output: Contribution to journalArticle

@article{c7d9592ce2df4f488144ae66b3592209,
title = "Neural quadratic discriminant analysis: Nonlinear decoding with V1-like computation",
abstract = "Linear-nonlinear (LN) models and their extensions have proven successful in describing transformations from stimuli to spiking responses of neurons in early stages of sensory hierarchies. Neural responses at later stages are highly nonlinear and have generally been better characterized in terms of their decoding performance on prespecified tasks. Here we develop a biologically plausible decoding model for classification tasks, that we refer to as neural quadratic discriminant analysis (nQDA). Specifically, we reformulate an optimal quadratic classifier as an LN-LN computation, analogous to {"}subunit{"} encoding models that have been used to describe responses in retina and primary visual cortex. We propose a physiological mechanism by which the parameters of the nQDA classifier could be optimized, using a supervised variant of a Hebbian learning rule. As an example of its applicability, we show that nQDA provides a better account than many comparable alternatives for the transformation between neural representations in two high-level brain areas recorded as monkeys performed a visual delayed-match-to-sample task.",
author = "Marino Pagan and Eero Simoncelli and Rust, {Nicole C.}",
year = "2016",
month = "11",
day = "1",
doi = "10.1162/NECO_a_00890",
language = "English (US)",
volume = "28",
pages = "2291--2319",
journal = "Neural Computation",
issn = "0899-7667",
publisher = "MIT Press Journals",
number = "11",

}

TY - JOUR

T1 - Neural quadratic discriminant analysis

T2 - Nonlinear decoding with V1-like computation

AU - Pagan, Marino

AU - Simoncelli, Eero

AU - Rust, Nicole C.

PY - 2016/11/1

Y1 - 2016/11/1

N2 - Linear-nonlinear (LN) models and their extensions have proven successful in describing transformations from stimuli to spiking responses of neurons in early stages of sensory hierarchies. Neural responses at later stages are highly nonlinear and have generally been better characterized in terms of their decoding performance on prespecified tasks. Here we develop a biologically plausible decoding model for classification tasks, that we refer to as neural quadratic discriminant analysis (nQDA). Specifically, we reformulate an optimal quadratic classifier as an LN-LN computation, analogous to "subunit" encoding models that have been used to describe responses in retina and primary visual cortex. We propose a physiological mechanism by which the parameters of the nQDA classifier could be optimized, using a supervised variant of a Hebbian learning rule. As an example of its applicability, we show that nQDA provides a better account than many comparable alternatives for the transformation between neural representations in two high-level brain areas recorded as monkeys performed a visual delayed-match-to-sample task.

AB - Linear-nonlinear (LN) models and their extensions have proven successful in describing transformations from stimuli to spiking responses of neurons in early stages of sensory hierarchies. Neural responses at later stages are highly nonlinear and have generally been better characterized in terms of their decoding performance on prespecified tasks. Here we develop a biologically plausible decoding model for classification tasks, that we refer to as neural quadratic discriminant analysis (nQDA). Specifically, we reformulate an optimal quadratic classifier as an LN-LN computation, analogous to "subunit" encoding models that have been used to describe responses in retina and primary visual cortex. We propose a physiological mechanism by which the parameters of the nQDA classifier could be optimized, using a supervised variant of a Hebbian learning rule. As an example of its applicability, we show that nQDA provides a better account than many comparable alternatives for the transformation between neural representations in two high-level brain areas recorded as monkeys performed a visual delayed-match-to-sample task.

UR - http://www.scopus.com/inward/record.url?scp=84994753608&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84994753608&partnerID=8YFLogxK

U2 - 10.1162/NECO_a_00890

DO - 10.1162/NECO_a_00890

M3 - Article

AN - SCOPUS:84994753608

VL - 28

SP - 2291

EP - 2319

JO - Neural Computation

JF - Neural Computation

SN - 0899-7667

IS - 11

ER -