特征选择的优化算法研究

目标识别成为战争胜败的关键,而对目标识别的关键之一是对目标特征的提取与选择.因此,特征的选择尤为重要.为了提高效率,通过一种算法选择较少（优化）的特征是所希望的.鉴于此该文简单介绍基于扩张矩阵与粗集理论的算法、启发式搜索算法、自适应神经网络、混沌神经网络等几种典型特征选择的优化算法的原理,并比较它们的性能,在此基础上提出了一种结合混沌神经网络和自适应神经网络的特征选择的改进方法,并对其原理进行简单介绍.最后用MATLAB编程验证启发式搜索算法特征选择的有效性.

Research on Optimal Algorithms of Feature Selection

Target recognition is vital for battle.And feature extraction and selection are vital for target recognition.So feature selection is especially impartant.Selecting relatively fewer features is hoped in order to enhance efficiency.Thus the paper introduces the theory of Extension Matrix and Rough Set,Heuristic search Algorithm,Adaptive Neural-Network,CNNA(chaos-neural networks algorithm) and so on and analyzes their capabilities.Based on this analogy,the paper presents an improved algorithm and analyzes its theory.The program of MATLAB proves the availability of heuristic search algorithm for selecting the optimal feature subset.