Structural learning and its applications to rule extraction
Masatoshi Ishikawa · 1994
The article presents the concept of structural learning with forgetting. To evaluate its effectiveness, various examples are tried, such as the discovery of a Boolean function and the classification of iris. Its extension to recurrent networks is also described. A database on mushrooms is used to demonstrate the effectiveness of rule extraction from training data. A comparative study on the performance of various structural learning methods is also reported.>