Paper Info
Title
Plasticity In The Granular Layer Enhances Motor Learning In A Computational Model Of The Cerebellum
Abstract
Learning mechanisms inspired by the animal cerebellum have shown promising achievements in artificial motor adaptation, mainly by focusing on the computation performed in the molecular layer. Other sites of cerebellar plasticity however are less explored whereas their understanding could contribute to improved computational solutions. In this study, we address the advantages of a form of plasticity found in the glomerulus, thought to control the temporal gating dynamics of the cerebellar pontine input. We explore this hypothesis from a system-level perspective within a simulated robotic rejection task, by implementing a model of the cerebellar microcircuit where adaptation of the input transformation dynamics, accounting for glomerular information processing, is controlled by a cost function. Our results suggest that glomerular adaptation (1) improves motor learning by adjusting input signal transformation properties towards an optimal configuration and shaping time and magnitude of the cerebellar output, and (2) contributes to fast readaptation during sudden plant perturbations. Finally, we discuss the implications of our results from a neuroscientific and articifical control perspective.
Year
DOI
Venue
2016
10.1007/978-3-319-44778-0_32
ARTIFICIAL NEURAL NETWORKS AND MACHINE LEARNING - ICANN 2016, PT I
Field
DocType
Volume
Granular layer,Climbing fiber,Neuroscience,Information processing,Gating,Motor learning,Computer science,Artificial intelligence,Automatic gain control,Cerebellum,Machine learning,Plasticity
Conference
9886
ISSN
Citations 
PageRank 
0302-9743
0
0.34
References 
Authors
3
4
Name
Order
Citations
PageRank
Giovanni Maffei1133.62
Ivan Herreros2358.13
Martí Sánchez-Fibla34411.13
Paul F. M. J. Verschure4677116.64