基于进化规划算法的舰船辐射噪声分类

神经网络分类器在舰船辐射噪声分类中得到了广泛的应用.针对神经网络分类器的设计困难,提出一种基于进化规划算法的设计方法.在该方法中,进化算法的适应度函数不是取为神经网络分类器对训练样本的识别率,而是对训练样本的可分性和聚合度同时考虑,这样能够在保证识别精度的前提下,使网络分类器具有良好的泛化能力,而且该方法不仅能够对待识别的样本进行离线学习,也能够在线学习.使用该分类器对舰船辐射噪声进行分类识别试验,结果表明该方法设计的分类器具有良好的性能.

Classification of Ship Radiation Noise Using Evolutionary Algorithm

Neural network classifier is widely used in ship radiation noise classification. According to the design difficulty of neural network classifier, an approach using evolutionary programming is proposed in the paper. Fitness function of evolutionary algorithm is not the classification rate of neural network but the combination of decentralization and conglomeration, then both the training set classification rate and testing set classification rate of neural network are high. Using this method, the neural netw...