免疫アルゴリズムをベースとするNSP専用エンジンアーキテクチャとその評価 Immune Algorithm Processor for Nurse Scheduling Problem
Recently, novel information processing mechanisms inspired by biological immune system are proposed and showed better capability than Neural Network and Genetic Algorithm. Biological immune system consists of high information processing mechanisms such as diverse antibody production mechanisms, self-regulating mechanism and primary and secondary immune responses based on antigen's specificity and immunological memory. Immune Algorithm (IA) based on this biological immune system can obtain multi optimal solutions without the constraint. However, in order to require very much computational time, it is not suitable for the actual problems. Therefore, the dedicated hardware is important in order to apply IA to actual problems. In this paper, we proposed Immune Algorithm processor for nurse scheduling problem. The proposed architecture is realized parallel processing on the calculation for affinity degree. Furthermore, the architecture is proposed not only the pipeline at evaluation phase, but also the pipeline on the whole at reconstruction phase. Experiment result evaluating the proposed architecture was shown to achieve more than 20 times the speed keeping the scheduling quality compared with software processing.
システム制御情報学会論文誌 19(8), 309-318, 2006-08-15