Peter Stone's Selected Publications

Classified by TopicClassified by Publication TypeSorted by DateSorted by First Author Last NameClassified by Funding Source


Machine Learning Capabilities of a Simulated Cerebellum

Matthew Hausknecht, Wen-Ke Li, Michael Mauk, and Peter Stone. Machine Learning Capabilities of a Simulated Cerebellum. "IEEE Transactions on Neural Networks and Learning Systems", 28(3):510–22, March 2017.

Download

[PDF]1.1MB  

Abstract

This article describes the learning and control capabilities of abiologically constrained bottom-up model of the mammaliancerebellum. Results are presented from six tasks - eyelidconditioning, pendulum balancing, PID control, robot balancing,pattern recognition, and MNIST handwritten digit recognition. Thesetasks span several paradigms of machine learning including supervisedlearning, reinforcement learning, control, and patternrecognition. Results over these six domains indicate that cerebellarsimulation is capable of robustly identifying static input patternseven when randomized across the sensory apparatus. This capabilityallows the simulated cerebellum to perform several differentsupervised learning and control tasks. On the other hand,reinforcement learning and temporal pattern recognition both proveproblematic due to the delayed nature of error signals and thesimulator's inability to solve the credit assignment problem. Theseresults are consistent with previous findings which hypothesize thatin the human brain, the basal ganglia is responsible for reinforcementlearning while the cerebellum handles supervised learning.

BibTeX Entry

@article{TNNLS15-hausknecht,
  author = {Matthew Hausknecht and Wen-Ke Li and Michael Mauk and Peter Stone},
  title = {Machine Learning Capabilities of a Simulated Cerebellum},
  Journal= {"IEEE Transactions on Neural Networks and Learning Systems"},
  Month="March",
  Year="2017",
  volume="28",
  number="3",
  pages="510--22",
  doi="10.1109/TNNLS.2015.2512838",
  abstract = {
This article describes the learning and control capabilities of a
biologically constrained bottom-up model of the mammalian
cerebellum. Results are presented from six tasks - eyelid
conditioning, pendulum balancing, PID control, robot balancing,
pattern recognition, and MNIST handwritten digit recognition. These
tasks span several paradigms of machine learning including supervised
learning, reinforcement learning, control, and pattern
recognition. Results over these six domains indicate that cerebellar
simulation is capable of robustly identifying static input patterns
even when randomized across the sensory apparatus. This capability
allows the simulated cerebellum to perform several different
supervised learning and control tasks. On the other hand,
reinforcement learning and temporal pattern recognition both prove
problematic due to the delayed nature of error signals and the
simulator's inability to solve the credit assignment problem. These
results are consistent with previous findings which hypothesize that
in the human brain, the basal ganglia is responsible for reinforcement
learning while the cerebellum handles supervised learning.
  },
}

Generated by bib2html.pl (written by Patrick Riley ) on Wed Nov 08, 2017 15:02:39