Using GPU to Shorten the Match Time of Rule Reasoning Based on Rete Algorithm

Usually, most of execution time of match-resolve-act reasoning cycle is spent in the matching phase. This issue has prevented the applicability of rule base systems. In this paper, the parallelism of α- and β-networks constructions in Rete algorithm have been realized on Graphics Processing Unit (GPU). It is possible to speed up the reasoning time 20 times faster than the current high performance multi-core processors. Furthermore, the parallel realization of Rete algorithm with GPU should be helpful on developing intelligent agents or data mining, using rule base systems.