基于拟随机序列与克隆选择的进化V-detector算法

Evolutionary V-detector algorithms based on clone selection and quasi random sequence

阴性选择(NS)算法是人工免疫的核心方法,检测器生成是具关键.针对具经典V-detector算法中高维数据失效及随机生成初始检测器集过于集中而导致过早收敛等问题,首先采用拟随机序列生成初始检测器;然后通过克降选择优化检测器集合,以覆盖非自体空问大小及数量作为亲和力标准,克服传统进化阴性选择(ENS)算法的局限性,并采用新型进化算子使得算法生成最优检测器集合;最后,通过实验验证了该方法的有效性.

Negative selection（NS） algorithm is the core algorithm of artificial immune system, in which the detector generate mechanism is the key. But the performance of V-detector algorithm becomes unfavorable on high-dimension data and the set of initial detectors randomly generated are too concentrated leading to the algorithm convergence prematurely. Quasi random sequence is used to generate the set of initial detectors. Then the detector set is optimized by using clone selection, and the coverage of non-self-space and the number of detectors are used as the standard of affinity which can over come the limitations of ENSA. A new selection, cloning and mutation operator is used to generate the optimal mature detector set. Finally, experiments verify the effectiveness of the proposed algorithm.