岭参数优化的ELM岭回归学习算法

ELM Ridge Regression Learning Algorithm of Ridge Parameter Optimization

针对ELM(extreme learning machine,极限学习机)学习算法可能存在的解的奇异问题,提出了岭参数优化的ELM岭回归学习算法(ELMRR).该算法利用岭回归方法代替原有的线性回归算法,以均方根误差为性能指标,采用粒子群优化算法确定最佳岭参数.为了验证该方法的有效性,对函数回归和分类问题进行仿真实验分析,结果表明该方法改善了ELM的预测性能且克服了传统岭回归算法岭参数难以确定的缺点.

Extreme learning machine ridge regression（ELMRR） learning algorithm of ridge parameter optimization is proposed to solve the problem that oddity solution possibly exists in ELM learning algorithm.The algorithm makes use of ridge regression instead of the previous linear regression,and uses particle swarm optimization algorithm to optimize ridge parameter according to root mean square error（RMSE）.Simulation experiment is performed for analyzing function regression and classification,and the effectiveness of this method is validated.The experimental results show that the algorithm improves predictive performance of ELM and overcomes the main flaw that it is difficult to obtain the ridge parameter in traditional ridge regression.