基于PSO的ELM集成学习算法研究

Research on Ensemble Learning Algorithm of ELM Based on PSO

极限学习机(Extreme Learning Machine,ELM)作为一种前向神经网络的学习算法,具有较快的学习速度和较高的学习精度。但是由于ELM的参数是随机设置的,导致ELM的泛化能力不高,容易产生过拟合现象。基于粒子群的极限学习机算法(PSO-ELM),提出了集成学习PSO-ELM算法,称为E-PSO-ELM,来提高模型的泛化能力。函数逼近和模式分类问题的仿真结果表明,所提出算法具有较好的泛化能力,特别是在函数逼近问题中,可以较为明显地提高模型的泛化能力和学习精度。

Extreme Learning Machine(ELM)is a forward neural networks learning algorithm with fast learning speed and learning accuracy.However,since the parameters of ELM are randomly set,so the generalization performance of ELM is poor,and over-fitting is likely to occur.Based on particle swarm optimization learning machine algorithm(PSO-ELM),we propose an ensemble PSO-ELM learning algorithm(called E-PSO-ELM)to improve the generalization ability of the model.The simulation results of function approximation and pattern classification show that the proposed algorithm has better generalization ability,especially in the function approximation problem,which can significantly improve the generalization ability and learning precision of the model.