基于主动进化的神经网络优化算法

摘要

The paper proposes an automatic optimization algorithm for neural networks based on active evolution and genetic algorithms. It is used to solve the network's parameters such as the weight and structural value. In order to solve the problem of low performance of genetic algorithms, the active evolution method is applied to the neural networks. The simulation result shows that the method has a higher convergence speed, improves the network's structural values and weight, and improves network's forecasting accuracy.%本文研究了一种基于主动进化神经网络的自动优化算法,用以同时完成对网络结构空间和权值空间的搜索.主动进化遗传算法在很大程度上克服了现有遗传算法执行效率低的问题,并因此运用到神经网络中去.实验结果表明,改进后的算法收敛速度更快,加快了网络拓扑结构和权值的搜寻速度,提高了网络预测精度.