Heuristic search and tree systems inference for structural pattern recognition

作者:

Highlights:

摘要

An experiment is reported that combines the use of a heuristic search and a tree systems inference technique to induce context-free relational grammars that generate patterns in the style embodied in a sample set. This is achieved by searching for concise structural representations of each pattern in a sample set and discovering regularities in their tree systems representation. The knowledge induced is represented as an accepting tree systems automaton and generative grammar, and is used to classify extant patterns and generate novel ones. Examples from the domain of architectural design are given to illustrate the methodology, and to lend support to the hypothesis that the style of a pattern can be partially characterized by a unique generative process.

论文关键词:grammatical inference,design description,p pattern recognition,pattern generation,structural recognition,machine learning

论文评审过程:Available online 19 February 2003.

论文官网地址:https://doi.org/10.1016/0950-7051(88)90002-0