Space coding for sensorimotor transformations can emerge through unsupervised learning

Michele De Filippo De Grazia, Simone Cutini, Matteo Lisi, Marco Zorzi

Research output: Contribution to journalArticle

7 Citations (Scopus)

Abstract

The posterior parietal cortex (PPC) is fundamental for sensorimotor transformations because it combines multiple sensory inputs and posture signals into different spatial reference frames that drive motor programming. Here, we present a computational model mimicking the sensorimotor transformations occurring in the PPC. A recurrent neural network with one layer of hidden neurons (restricted Boltzmann machine) learned a stochastic generative model of the sensory data without supervision. After the unsupervised learning phase, the activity of the hidden neurons was used to compute a motor program (a population code on a bidimensional map) through a simple linear projection and delta rule learning. The average motor error, calculated as the difference between the expected and the computed output, was less than 3. Importantly, analyses of the hidden neurons revealed gain-modulated visual receptive fields, thereby showing that space coding for sensorimotor transformations similar to that observed in the PPC can emerge through unsupervised learning. These results suggest that gain modulation is an efficient coding strategy to integrate visual and postural information toward the generation of motor commands.

Original languageEnglish
JournalCognitive Processing
Volume13
Issue number1 SUPPL
DOIs
Publication statusPublished - 2012

Fingerprint

Parietal Lobe
Unsupervised learning
Learning
Neurons
Population Control
Visual Fields
Posture
Recurrent neural networks
Stochastic models
Modulation

Keywords

  • Gain modulation
  • Generative model
  • Neural network
  • Parietal cortex
  • Sensorimotor transformations

ASJC Scopus subject areas

  • Cognitive Neuroscience
  • Experimental and Cognitive Psychology
  • Artificial Intelligence

Cite this

Space coding for sensorimotor transformations can emerge through unsupervised learning. / De Filippo De Grazia, Michele; Cutini, Simone; Lisi, Matteo; Zorzi, Marco.

In: Cognitive Processing, Vol. 13, No. 1 SUPPL, 2012.

Research output: Contribution to journalArticle

De Filippo De Grazia, Michele ; Cutini, Simone ; Lisi, Matteo ; Zorzi, Marco. / Space coding for sensorimotor transformations can emerge through unsupervised learning. In: Cognitive Processing. 2012 ; Vol. 13, No. 1 SUPPL.
@article{939006926c6c4a52a8952b39de377590,
title = "Space coding for sensorimotor transformations can emerge through unsupervised learning",
abstract = "The posterior parietal cortex (PPC) is fundamental for sensorimotor transformations because it combines multiple sensory inputs and posture signals into different spatial reference frames that drive motor programming. Here, we present a computational model mimicking the sensorimotor transformations occurring in the PPC. A recurrent neural network with one layer of hidden neurons (restricted Boltzmann machine) learned a stochastic generative model of the sensory data without supervision. After the unsupervised learning phase, the activity of the hidden neurons was used to compute a motor program (a population code on a bidimensional map) through a simple linear projection and delta rule learning. The average motor error, calculated as the difference between the expected and the computed output, was less than 3. Importantly, analyses of the hidden neurons revealed gain-modulated visual receptive fields, thereby showing that space coding for sensorimotor transformations similar to that observed in the PPC can emerge through unsupervised learning. These results suggest that gain modulation is an efficient coding strategy to integrate visual and postural information toward the generation of motor commands.",
keywords = "Gain modulation, Generative model, Neural network, Parietal cortex, Sensorimotor transformations",
author = "{De Filippo De Grazia}, Michele and Simone Cutini and Matteo Lisi and Marco Zorzi",
year = "2012",
doi = "10.1007/s10339-012-0478-4",
language = "English",
volume = "13",
journal = "Cognitive Processing",
issn = "1612-4782",
publisher = "Springer Verlag",
number = "1 SUPPL",

}

TY - JOUR

T1 - Space coding for sensorimotor transformations can emerge through unsupervised learning

AU - De Filippo De Grazia, Michele

AU - Cutini, Simone

AU - Lisi, Matteo

AU - Zorzi, Marco

PY - 2012

Y1 - 2012

N2 - The posterior parietal cortex (PPC) is fundamental for sensorimotor transformations because it combines multiple sensory inputs and posture signals into different spatial reference frames that drive motor programming. Here, we present a computational model mimicking the sensorimotor transformations occurring in the PPC. A recurrent neural network with one layer of hidden neurons (restricted Boltzmann machine) learned a stochastic generative model of the sensory data without supervision. After the unsupervised learning phase, the activity of the hidden neurons was used to compute a motor program (a population code on a bidimensional map) through a simple linear projection and delta rule learning. The average motor error, calculated as the difference between the expected and the computed output, was less than 3. Importantly, analyses of the hidden neurons revealed gain-modulated visual receptive fields, thereby showing that space coding for sensorimotor transformations similar to that observed in the PPC can emerge through unsupervised learning. These results suggest that gain modulation is an efficient coding strategy to integrate visual and postural information toward the generation of motor commands.

AB - The posterior parietal cortex (PPC) is fundamental for sensorimotor transformations because it combines multiple sensory inputs and posture signals into different spatial reference frames that drive motor programming. Here, we present a computational model mimicking the sensorimotor transformations occurring in the PPC. A recurrent neural network with one layer of hidden neurons (restricted Boltzmann machine) learned a stochastic generative model of the sensory data without supervision. After the unsupervised learning phase, the activity of the hidden neurons was used to compute a motor program (a population code on a bidimensional map) through a simple linear projection and delta rule learning. The average motor error, calculated as the difference between the expected and the computed output, was less than 3. Importantly, analyses of the hidden neurons revealed gain-modulated visual receptive fields, thereby showing that space coding for sensorimotor transformations similar to that observed in the PPC can emerge through unsupervised learning. These results suggest that gain modulation is an efficient coding strategy to integrate visual and postural information toward the generation of motor commands.

KW - Gain modulation

KW - Generative model

KW - Neural network

KW - Parietal cortex

KW - Sensorimotor transformations

UR - http://www.scopus.com/inward/record.url?scp=84872812480&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84872812480&partnerID=8YFLogxK

U2 - 10.1007/s10339-012-0478-4

DO - 10.1007/s10339-012-0478-4

M3 - Article

VL - 13

JO - Cognitive Processing

JF - Cognitive Processing

SN - 1612-4782

IS - 1 SUPPL

ER -