Effective Reinforcement learning following cerebellar damage requires a balance between exploration and motor noise

Amanda S. Therrien, Daniel M. Wolpert, Amy J Bastian

Research output: Contribution to journalArticle

Abstract

Reinforcement and error-based processes are essential for motor learning, with the cerebellum thought to be required only for the error-based mechanism. Here we examined learning and retention of a reaching skill under both processes. Control subjects learned similarly from reinforcement and error-based feedback, but showed much better retention under reinforcement. To apply reinforcement to cerebellar patients, we developed a closed-loop reinforcement schedule in which task difficulty was controlled based on recent performance. This schedule produced substantial learning in cerebellar patients and controls. Cerebellar patients varied in their learning under reinforcement but fully retained what was learned. In contrast, they showed complete lack of retention in errorbased learning. We developed a mechanistic model of the reinforcement task and found that learning depended on a balance between exploration variability and motor noise. While the cerebellar and control groups had similar exploration variability, the patients had greater motor noise and hence learned less. Our results suggest that cerebellar damage indirectly impairs reinforcement learning by increasing motor noise, but does not interfere with the reinforcement mechanism itself. Therefore, reinforcement can be used to learn and retain novel skills, but optimal reinforcement learning requires a balance between exploration variability and motor noise.

Original languageEnglish (US)
Pages (from-to)101-114
Number of pages14
JournalBrain
Volume139
Issue number1
DOIs
StatePublished - Jan 1 2016

Fingerprint

Noise
Learning
Reinforcement (Psychology)
Reinforcement Schedule
Cerebellum
Appointments and Schedules
Control Groups

Keywords

  • Adaptation
  • Ataxia
  • Cerebellum
  • Reinforcement learning
  • Visuomotor rotation

ASJC Scopus subject areas

  • Clinical Neurology

Cite this

Effective Reinforcement learning following cerebellar damage requires a balance between exploration and motor noise. / Therrien, Amanda S.; Wolpert, Daniel M.; Bastian, Amy J.

In: Brain, Vol. 139, No. 1, 01.01.2016, p. 101-114.

Research output: Contribution to journalArticle

@article{5e67673247934f879400d1dae4c95971,
title = "Effective Reinforcement learning following cerebellar damage requires a balance between exploration and motor noise",
abstract = "Reinforcement and error-based processes are essential for motor learning, with the cerebellum thought to be required only for the error-based mechanism. Here we examined learning and retention of a reaching skill under both processes. Control subjects learned similarly from reinforcement and error-based feedback, but showed much better retention under reinforcement. To apply reinforcement to cerebellar patients, we developed a closed-loop reinforcement schedule in which task difficulty was controlled based on recent performance. This schedule produced substantial learning in cerebellar patients and controls. Cerebellar patients varied in their learning under reinforcement but fully retained what was learned. In contrast, they showed complete lack of retention in errorbased learning. We developed a mechanistic model of the reinforcement task and found that learning depended on a balance between exploration variability and motor noise. While the cerebellar and control groups had similar exploration variability, the patients had greater motor noise and hence learned less. Our results suggest that cerebellar damage indirectly impairs reinforcement learning by increasing motor noise, but does not interfere with the reinforcement mechanism itself. Therefore, reinforcement can be used to learn and retain novel skills, but optimal reinforcement learning requires a balance between exploration variability and motor noise.",
keywords = "Adaptation, Ataxia, Cerebellum, Reinforcement learning, Visuomotor rotation",
author = "Therrien, {Amanda S.} and Wolpert, {Daniel M.} and Bastian, {Amy J}",
year = "2016",
month = "1",
day = "1",
doi = "10.1093/brain/awv329",
language = "English (US)",
volume = "139",
pages = "101--114",
journal = "Brain",
issn = "0006-8950",
publisher = "Oxford University Press",
number = "1",

}

TY - JOUR

T1 - Effective Reinforcement learning following cerebellar damage requires a balance between exploration and motor noise

AU - Therrien, Amanda S.

AU - Wolpert, Daniel M.

AU - Bastian, Amy J

PY - 2016/1/1

Y1 - 2016/1/1

N2 - Reinforcement and error-based processes are essential for motor learning, with the cerebellum thought to be required only for the error-based mechanism. Here we examined learning and retention of a reaching skill under both processes. Control subjects learned similarly from reinforcement and error-based feedback, but showed much better retention under reinforcement. To apply reinforcement to cerebellar patients, we developed a closed-loop reinforcement schedule in which task difficulty was controlled based on recent performance. This schedule produced substantial learning in cerebellar patients and controls. Cerebellar patients varied in their learning under reinforcement but fully retained what was learned. In contrast, they showed complete lack of retention in errorbased learning. We developed a mechanistic model of the reinforcement task and found that learning depended on a balance between exploration variability and motor noise. While the cerebellar and control groups had similar exploration variability, the patients had greater motor noise and hence learned less. Our results suggest that cerebellar damage indirectly impairs reinforcement learning by increasing motor noise, but does not interfere with the reinforcement mechanism itself. Therefore, reinforcement can be used to learn and retain novel skills, but optimal reinforcement learning requires a balance between exploration variability and motor noise.

AB - Reinforcement and error-based processes are essential for motor learning, with the cerebellum thought to be required only for the error-based mechanism. Here we examined learning and retention of a reaching skill under both processes. Control subjects learned similarly from reinforcement and error-based feedback, but showed much better retention under reinforcement. To apply reinforcement to cerebellar patients, we developed a closed-loop reinforcement schedule in which task difficulty was controlled based on recent performance. This schedule produced substantial learning in cerebellar patients and controls. Cerebellar patients varied in their learning under reinforcement but fully retained what was learned. In contrast, they showed complete lack of retention in errorbased learning. We developed a mechanistic model of the reinforcement task and found that learning depended on a balance between exploration variability and motor noise. While the cerebellar and control groups had similar exploration variability, the patients had greater motor noise and hence learned less. Our results suggest that cerebellar damage indirectly impairs reinforcement learning by increasing motor noise, but does not interfere with the reinforcement mechanism itself. Therefore, reinforcement can be used to learn and retain novel skills, but optimal reinforcement learning requires a balance between exploration variability and motor noise.

KW - Adaptation

KW - Ataxia

KW - Cerebellum

KW - Reinforcement learning

KW - Visuomotor rotation

UR - http://www.scopus.com/inward/record.url?scp=84964612276&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84964612276&partnerID=8YFLogxK

U2 - 10.1093/brain/awv329

DO - 10.1093/brain/awv329

M3 - Article

C2 - 26626368

AN - SCOPUS:84964612276

VL - 139

SP - 101

EP - 114

JO - Brain

JF - Brain

SN - 0006-8950

IS - 1

ER -