基于遗传算法的神经网络优化

Artificial neural networks optimizing on genetic algorithm

针对神经网络结构的设计基本上依赖于人的经验的缺点,提出了一种利用遗传算法优化神经网络的算法。该算法结合了神经网络的快速并行性和遗传算法的全局搜索性,对神经网络的连接权和结构进行了优化,剔除整个网络冗余节点和冗余连接权,提高了网络的处理能力。通过实例证明该算法具有很高的精度,在机械故障诊断中具有良好的应用前景。

In case of the structure of artificial neural networks was decided by one’s experience, an optimized algorithm of artificial neural networks by genetic algorithm was proposed in this paper. In the algorithm, the global property of genetic algorithm (GA) and the parallelisms of artificial neural networks (ANN) were combined; weights and structure of artificial neural networks were optimized together. Raised the processing ability of networks, the redundant nodes and weights were eliminated. The illustrational result shows that the algorithm can improve the accuracy and has the bright future in the application of machinery fault diagnosis.