事例ベース推論とグラフ構造の弁別ネット

概要

論文の詳細を見る
Case-based reasoning (CBR) is expected to be the architecture of the next generation expert systems. Discrimination net of graphs, which has the capability of similarity search and analogical reasoning in chemistry domain, shares many characteristics with case based reasoners. Moreover, discrimination net method has an advantage that no knowledge acquisition of domain specific rules are necessary but the meta-level reasoning method provided can be an inference tool for a variety of problems. Both conventional CBR and discrimination net methods are briefly introduced, and they are compared to each other with respect to their process models and memory organizations. Extension of the discrimination net architecture is discussed to treat the problems other than chemistry domain, and proper extension of the method is shown to be useful as a general architecture of CBR expert shells.
関西学院大学の論文
1992-07-01