论文部分内容阅读
针对极端学习机(ELM)网络结构设计问题,提出基于灵敏度分析法的ELM剪枝算法.利用隐含层节点输出和相对应的输出层权值向量,定义学习残差对于隐含层节点的灵敏度和网络规模适应度,根据灵敏度大小判断隐含层节点的重要性,利用网络规模适应度确定隐含层节点个数,删除重要性较低的节点.仿真结果表明,所提出的算法能够较为准确地确定与学习样本相匹配的网络规模,解决了ELM网络结构设计问题.
In order to solve the problem of ELM network structure design, an ELM pruning algorithm based on sensitivity analysis is proposed.Using the output of hidden layer node and the corresponding output layer weight vector, the sensitivity of learning residual to hidden layer node is defined And the network scale fitness, the importance of hidden layer nodes is judged according to the sensitivity, the number of hidden layer nodes is determined by using the network scale fitness, and the less important nodes are deleted. The simulation results show that the proposed algorithm can be more accurate Determine the size of the network that matches the learning sample, and solve the ELM network structure design problem.