The subsumption mechanism for XCS using code fragmented conditions
Hsuan-Ta Lin, Po-Ming Lee, Tzu‐Chien Hsiao · 2013
By utilizing a code-fragmented representation of Extended Classifier System (XCS) condition in conjunction with building-block extraction technique, autonomous scaling has been realized in the latest work of XCS. The technique substantially reduces the number of training instances required in various benchmark problems. However, the subsumption mechanism was not included in the former report of the technique. Therefore, we invented the subsumption mechanism for XCS with such technique, and observed the characteristics of such the system in multiplexer problems. The finding indicates that our subsumption mechanism decreased the number of macro-classifiers.