Basal Ganglia Models for Autonomous Behavior Learning

Authors:
Hiroshi Tsujino;Johane Takeuchi;Osamu Shouno
Affiliations:
Honda Research Institute Japan Co., Ltd., Saitama, Japan 351-0188;Honda Research Institute Japan Co., Ltd., Saitama, Japan 351-0188;Honda Research Institute Japan Co., Ltd., Saitama, Japan 351-0188
Venue:
Creating Brain-Like Intelligence
Year:
2009

Citing 11
Cited 1

Integrated architecture for learning, planning, and reacting based on approximating dynamic programming

Proceedings of the seventh international conference (1990) on Machine learning
Technical Note: \cal Q-Learning

Machine Learning
A cortical-type modular neural network for hypothetical reasoning

Neural Networks
What are the computations of the cerebellum, the basal ganglia and the cerebral cortex?

Neural Networks - Special issue on organisation of computation in brain-like systems
Introduction to Reinforcement Learning

Introduction to Reinforcement Learning
Multiple model-based reinforcement learning

Neural Computation
Learning to Predict by the Methods of Temporal Differences

Machine Learning
The Basal Ganglia and Cortex Implement Optimal Decision Making Between Alternative Actions

Neural Computation
Learning to act using real-time dynamic programming

Artificial Intelligence
An online adaptation control system using mnSOM

ICONIP'06 Proceedings of the 13 international conference on Neural Information Processing - Volume Part I
Intrinsic Motivation Systems for Autonomous Mental Development

IEEE Transactions on Evolutionary Computation

Creating Brain-Like Intelligence

Creating Brain-Like Intelligence

Quantified Score

Hi-index	0.00

Visualization

Abstract

We propose two basal ganglia (BG) models for autonomous behavior learning: the BG system model and the BG spiking neural network model. These models were developed on the basis of reinforcement learning (RL) theories and neuroscience principals of behavioral learning. The BG system model focuses on problems with RL input selection and reward setting. This model assumes that parallel BG modules receive a variety of inputs. We also propose an automatic setting method of internal reward for this model. The BG spiking neural network model focuses on problems with biological neural network architecture, ambiguous inputs and the mechanism of timing. This model accounts for the neurophysiological characteristics of neurons and differential functions of the direct and indirect pathways. We demonstrate that the BG system model achieves goals in fewer trials by learning the internal state representation, whereas the BG spiking neural network model has the capacity for probabilistic selection of action. Our results suggest that these two models are a step toward developing an autonomous learning system.