基于增强随机搜索的OECI-ELM算法

摘要

为提高 OCI-ELM算法性能,从神经网络结构设计出发,将 EI-ELM中增强随机搜索能力的机制应用于 OCI-ELM 算法中,提出一种增长结构的新算法(Optimal Incremental Coeffi-cients-Extreme Learning Machine,OECI-ELM)。数值实验表明,由该算法得到的网络序列可以逼近任意连续的目标函数。对于不同的回归问题,与 OCI-ELM、EI-ELM、CI-ELM、ECI-ELM算法相比,OECI-ELM算法可获得更加紧凑的网络结构,且泛化能力也更强。%To improve optimal incremental coefficients extreme learning machine,the OECI-ELM (Op-timal Incremental Coefficients Extreme Learning Machine)is proposed by introducing the random search mechanism of EI-ELM into OCI-ELM,based on neural network structure design.The simula-tion results show that OECI-ELM can approximate any continuous target function.For the various re-gression problems,OECI-ELM can obtain a more compact network architecture and better generaliza-tion performance,compared with OCI-ELM,EI-ELM,CI-ELM and ECI-ELM.