摘要 |
<P>PROBLEM TO BE SOLVED: To estimate a feature pattern corresponding to an observation from a pattern of an observation target at extremely reduced calculation costs. <P>SOLUTION: In order to learn a correlation between an inputted observation pattern and a pattern of an estimation target in which an output desired to find is estimated from the input, a first teaching vector composed of the observation pattern and a second teaching vector composed of the estimation target are inputted and while calculating self-correlation information of the two teaching vectors and mutual correlation information of an average vector, the first teaching vector and the second teaching vector, by using such information, a stochastic expected value based on the Bayesian theory of the estimation target is found with respect to the input pattern so that the desired parameter is accurately estimated at the low calculation costs. <P>COPYRIGHT: (C)2004,JPO |