免疫进化否定选择算法

摘要

当训练样本分布密集交错时,传统的否定选择算法难以将检测器生成在正/反样本间的有效区域,导致检测器集合对这些样本的识别率降低,影响了算法性能.为使检测器能有效地识别分布密集交错的样本,提出了免疫进化否定选择算法(IENSA).IENSA通过加入两个免疫进化过程,首先在样本分布密集的区域引导检测器在正/反样本之间有效地生成,然后在样本分布稀疏的区域对冗余检测器进行抑制.实验结果表明在二维人工数据集Rectangle与三维标准数据集Skin segmentation上,相对于经典的RNSA与V-detector算法,IENSA均能以较少的检测器达到较高的检测率.%When the samples distribute densely,the traditional negative selection algorithm is difficult to generate detectors in the gap between normal and abnormal samples,it causes that the algorithm has the low detecting rate for these samples.In order to enable the detector to effectively identify the densely samples,this paper proposed the immune evolution negative selection algorithm (IENSA).By adding two immune evolution processes,IENSA could generate detector in the gap between normal and abnormal samples effectively,and restrain the redundant detector in the sparse area of the sample distribution.The experimental result show that,on the artificial data set Rectangle (2D) and the UCI standard data set Skin segmentation (3D),compared to the classical RNSA and V-detector algorithm,IENSA can reach the higher detection rate with the less antibodies and training time.