CLIP: concept learning from inference patterns

摘要: Abstract A new concept-learning method called CLIP (concept learning from inference patterns) is proposed that learns concepts patterns, not positive/negative examples most conventional concept methods use. The learned enable an efficient on a more abstract level. We use colored digraph to represent patterns. graph representation expressive enough and enables the quantitative analysis of pattern frequency. process consists following two steps: (1) Convert original patterns digraph, (2) Extract set typical which appears frequently in digraph. basic idea smaller becomes, amount data be handled becomes and, accordingly, uses these data. Also, we can reduce size by replacing each appearing with single node, reduced node represents concept. Experimentally, automatically generates multilevel representations given physical/single-level carry-chain circuit. These involve descriptions circuit, such as mathematical logical descriptions.