一种新的混合智能极限学习机

An improved hybrid intelligent extreme learning machine

提出一种基于差分进化(DE)和粒子群优化(PSO)的混合智能方法——DEPSO算法,并通过对10个典型函数进行测试,表明DEPSO算法具有良好的寻优性能.针对单隐层前向神经网络(SLFNs)提出一种改进的学习算法—DEPSO-ELM算法,即应用DEPSO算法优化SLFNs的隐层节点参数,采用极限学习算法(ELM)求取SLFNs的输出权值.将DEPSO-ELM算法应用于6个典型真实数据集的回归计算,并与DE-ELM、SaE-ELM算法相比,获得了更精确的计算结果.最后,将DEPSO-ELM算法应用于数控机床热误差的建模预测,获得了良好的预测效果.

An improved hybrid intelligent algorithm based on differential evolution（DE） and particle swarm optimization （PSO） is proposed. The performance of DEPSO algorithm is verified by simulations on 10 benchmark functions. Then, an improved learning algorithm named DEPSO extreme learning machine（DEPSO-ELM） algorithm for single hidden layer feedforward networks（SLFNs） is proposed. In DEPSO-ELM, DEPSO is used to optimize the network hidden node parameters, and ELM is used to analytically determine the output weights. Simulation results of 6 real world datasets regression problems show that the DEPSO-ELM algorithm performs better than DE-ELM and SaE-ELM. Finally, the effectiveness of the DEPSO-ELM algorithm is verified in the prediction of NC machine tool thermal errors.