Humans can generate accurate and appropriate motor commands in various and even uncertain environments. MOSAIC (MOdular Sellection And Identification for Control) was formerly proposed for describing such human ability, but it includes some complex and heuristic procedures which make the model's understandability hard. In this article, we present an alternative and probabilistic model of MOSAIC (p-MOSAIC) as a mixture of normal distributions, and an online EM-based learning method for its predictors and controllers. Theoretical consideration shows that the learning rule of p-MOSAIC corresponds to that of MOSAIC except for some points mostly related to the controller learning. Experimental studies using synthetic datasets have shown some practical advantages of p-MOSAIC. One is that the learning rule of p-MOSAIC makes the estimation of 'responsibility' stable. Another is that p-MOSAIC realizes accurate control and robust parameter learning in comparison to the original MOSAIC especially in noisy environments, due to the direct incorporation of the noise into the model
[1]
Shin Ishii,et al.
On-line EM Algorithm for the Normalized Gaussian Network
,
2000,
Neural Computation.
[2]
Michael I. Jordan.
Learning in Graphical Models
,
1999,
NATO ASI Series.
[3]
Heekuck Oh,et al.
Neural Networks for Pattern Recognition
,
1993,
Adv. Comput..
[4]
B. Bridgeman,et al.
Alternating prism exposure causes dual adaptation and generalization to a novel displacement
,
1993,
Perception & psychophysics.
[5]
Mitsuo Kawato,et al.
Feedback-Error-Learning Neural Network for Supervised Motor Learning
,
1990
.
[6]
Mitsuo Kawato,et al.
MOSAIC Model for Sensorimotor Learning and Control
,
2001,
Neural Computation.
[7]
Arun D Kulkarni,et al.
Neural Networks for Pattern Recognition
,
1991
.
[8]
D M Wolpert,et al.
Multiple paired forward and inverse models for motor control
,
1998,
Neural Networks.