摘要 |
<P>PROBLEM TO BE SOLVED: To solve the problem of a conventional learning system of a neutral network that, while learning is done using a single neutral network for data to be classified, the classification capability is reduced if there exists a case in which noise or contradictory data is included in a database. <P>SOLUTION: A database analyzing system using a neutral network according to the present invention divides the database into optimum subspaces by using a plurality of learning means of neutral networks, thereby enabling specialized classification for the respective subspaces. <P>COPYRIGHT: (C)2005,JPO&NCIPI |