基于结构自适应神经网络特征选择的一种改进方法

对电子目标识别的关键之一是对目标特征的提取与选择。首先介绍两种特征选择算法:结构自适应神经网络和混沌神经网络,然后对它们的性能进行比较,并分析它们各自的优缺点,在此基础上提出一种特征选择的改进方法,即结合混沌神经网络和结构自适应神经网络的特征选择法,并通过仿真和理论分析验证该方法的可行性。

A modified feature selection approach based on structure-adaptive neural-networks

Feature extraction and selection is vital for target recognition.First,the theory and structure of two methods for feature selection,Adaptive-Neural-Network and Chaos-Neural-Networks(CNN),are introduced.Then comparison is made between their capabilities,and the respective advantages and disadvantages of each method are analyzed.On the basis of which the authors put forward a modified method for feature selection,i.e.combining Adaptive-Neural-Networks with CNN.Simulations and theoretical analysis proved the feasibility of this method.