一种黄金分割优化的极限学习机算法

针对极限学习机(ELM)存在大量隐层神经元个数和随机给定权值导致算法性能不稳定等问题,将黄金分割法(Golden Section)与ELM相结合提出了基于黄金分割优化的极限学习机算法(GS-ELM).首先通过黄金分割法对ELM隐含层节点数进行优化,接着再用该方法对ELM输入层权值和隐含层偏差进行优化.实验结果表明,相比较传统的BP神经网络,支持向量机和极限学习机,GS-ELM算法能获得较高的分类精度.

A new extreme learning machine based on the golden section optimization algorithm

In view of the problems of a large number of hidden layer neuron nodes and the instability of the algorithm performance caused by random weights of the extreme learning machine （ELM）,an improved extreme learning machine algorithm based on golden section optimization （GS-ELM） was proposed.The golden section optimization algorithm was used to optimize the number of hidden layer nodes,the weights of the input layers and the bias of the hidden layers.The experimental results showed that the GS-ELM algorithm can achieve higher classification accuracy compared with the traditional BP neural network,support vector machine （SVM） and the extreme learning machine.