基于量子遗传算法的特征选择算法

特征选择是模式识别和机器学习等领域中重要而困难的研究课题.提出一种最优特征子集评价准则和实现特征选择的一种新量子遗传算法(NQGA).NQGA采用量子门旋转角更新新方法和增强算法寻优能力及防止早熟收敛的移民和灾变策略.定性分析了NQGA的高效性.典型复杂函数测试和雷达辐射源信号特征选择的应用表明,NQGA寻优能力强、收敛速度快和能有效防止早熟现象.采用提出的准则函数和搜索策略实现特征选择,大大降低了特征维数,获得了更高的正确识别率.

Feature selection algorithm based on quantum genetic algorithm

Feature selection is always an important and difficult issue in pattern recognition and machine learning.This paper proposed a criterion function for selecting the optimal feature subset and a search strategy called novel quantum genetic algorithm(NQGA).NQGA adopted a novel update approach of rotation angles of quantum gates,and immigration and catastrophe operations to enhance search capability and to avoid premature convergence.Besides,high efficiency of NQGA was analyzed qualitatively.Testing results of typically complex functions and experimental results of feature selection in radar emitter signal recognition show that NQGA has good characteristics of strong search capability,rapid convergence and no premature convergence.The proposed feature selection algorithm reduces greatly the dimensions of original feature set and heightens accurate recognition rate of radar emitter signals.