Inference in the Brain: Statistics Flowing in Redundant Population Codes

Xaq Pitkow, Dora Angelaki

Research output: Contribution to journalReview article

Abstract

It is widely believed that the brain performs approximate probabilistic inference to estimate causal variables in the world from ambiguous sensory data. To understand these computations, we need to analyze how information is represented and transformed by the actions of nonlinear recurrent neural networks. We propose that these probabilistic computations function by a message-passing algorithm operating at the level of redundant neural populations. To explain this framework, we review its underlying concepts, including graphical models, sufficient statistics, and message-passing, and then describe how these concepts could be implemented by recurrently connected probabilistic population codes. The relevant information flow in these networks will be most interpretable at the population level, particularly for redundant neural codes. We therefore outline a general approach to identify the essential features of a neural message-passing algorithm. Finally, we argue that to reveal the most important aspects of these neural computations, we must study large-scale activity patterns during moderately complex, naturalistic behaviors.

Original languageEnglish (US)
Pages (from-to)943-953
Number of pages11
JournalNeuron
Volume94
Issue number5
DOIs
StatePublished - Jun 7 2017

Fingerprint

Brain
Population

Keywords

  • brain
  • coding
  • inference
  • message-passing
  • nonlinear
  • nuisance
  • population code
  • redundant
  • theory

ASJC Scopus subject areas

  • Neuroscience(all)

Cite this

Inference in the Brain : Statistics Flowing in Redundant Population Codes. / Pitkow, Xaq; Angelaki, Dora.

In: Neuron, Vol. 94, No. 5, 07.06.2017, p. 943-953.

Research output: Contribution to journalReview article

@article{e3dd4c1dd85142ae8b1852a7b59ce082,
title = "Inference in the Brain: Statistics Flowing in Redundant Population Codes",
abstract = "It is widely believed that the brain performs approximate probabilistic inference to estimate causal variables in the world from ambiguous sensory data. To understand these computations, we need to analyze how information is represented and transformed by the actions of nonlinear recurrent neural networks. We propose that these probabilistic computations function by a message-passing algorithm operating at the level of redundant neural populations. To explain this framework, we review its underlying concepts, including graphical models, sufficient statistics, and message-passing, and then describe how these concepts could be implemented by recurrently connected probabilistic population codes. The relevant information flow in these networks will be most interpretable at the population level, particularly for redundant neural codes. We therefore outline a general approach to identify the essential features of a neural message-passing algorithm. Finally, we argue that to reveal the most important aspects of these neural computations, we must study large-scale activity patterns during moderately complex, naturalistic behaviors.",
keywords = "brain, coding, inference, message-passing, nonlinear, nuisance, population code, redundant, theory",
author = "Xaq Pitkow and Dora Angelaki",
year = "2017",
month = "6",
day = "7",
doi = "10.1016/j.neuron.2017.05.028",
language = "English (US)",
volume = "94",
pages = "943--953",
journal = "Neuron",
issn = "0896-6273",
publisher = "Cell Press",
number = "5",

}

TY - JOUR

T1 - Inference in the Brain

T2 - Statistics Flowing in Redundant Population Codes

AU - Pitkow, Xaq

AU - Angelaki, Dora

PY - 2017/6/7

Y1 - 2017/6/7

N2 - It is widely believed that the brain performs approximate probabilistic inference to estimate causal variables in the world from ambiguous sensory data. To understand these computations, we need to analyze how information is represented and transformed by the actions of nonlinear recurrent neural networks. We propose that these probabilistic computations function by a message-passing algorithm operating at the level of redundant neural populations. To explain this framework, we review its underlying concepts, including graphical models, sufficient statistics, and message-passing, and then describe how these concepts could be implemented by recurrently connected probabilistic population codes. The relevant information flow in these networks will be most interpretable at the population level, particularly for redundant neural codes. We therefore outline a general approach to identify the essential features of a neural message-passing algorithm. Finally, we argue that to reveal the most important aspects of these neural computations, we must study large-scale activity patterns during moderately complex, naturalistic behaviors.

AB - It is widely believed that the brain performs approximate probabilistic inference to estimate causal variables in the world from ambiguous sensory data. To understand these computations, we need to analyze how information is represented and transformed by the actions of nonlinear recurrent neural networks. We propose that these probabilistic computations function by a message-passing algorithm operating at the level of redundant neural populations. To explain this framework, we review its underlying concepts, including graphical models, sufficient statistics, and message-passing, and then describe how these concepts could be implemented by recurrently connected probabilistic population codes. The relevant information flow in these networks will be most interpretable at the population level, particularly for redundant neural codes. We therefore outline a general approach to identify the essential features of a neural message-passing algorithm. Finally, we argue that to reveal the most important aspects of these neural computations, we must study large-scale activity patterns during moderately complex, naturalistic behaviors.

KW - brain

KW - coding

KW - inference

KW - message-passing

KW - nonlinear

KW - nuisance

KW - population code

KW - redundant

KW - theory

UR - http://www.scopus.com/inward/record.url?scp=85020250628&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85020250628&partnerID=8YFLogxK

U2 - 10.1016/j.neuron.2017.05.028

DO - 10.1016/j.neuron.2017.05.028

M3 - Review article

VL - 94

SP - 943

EP - 953

JO - Neuron

JF - Neuron

SN - 0896-6273

IS - 5

ER -