结合期望风险的极限学习机的研究

对极限学习机的模型进行了研究，提出了一种结合期望风险最小化的极限学习机的预测模型。其基本思想是同时考虑结构风险和期望风险，根据期望风险和经验风险之间的关系，将期望风险转换成经验风险，进行最小化期望风险的极限学习机预测模型求解。利用人工数据集和实际数据集进行回归问题的数值实验，并与极限学习机（ELM）和正则极限学习机（RELM）两种算法的性能进行了比较，实验结果表明，所提方法能有效提高了泛化能力。

Research on extreme learning machine with expected risk

Research on the model of extreme learning machine, a prediction model is proposed of extreme learning machine which is based on expected risk minimization. It’s basic idea is to consider both structure risk and expected risk at the same time, according to relationship between expected risk and empirical risk, converting expected risk into empirical risk, so that prediction model of extreme learning machine can be solve with minimizing expected risk. Using artificial data set and real data set of regression results, and compared with Extreme Learning Machine （ELM） and Regular Extreme Learning Machine （RELM） two kinds of algorithm performance. Experimental results show that the proposed method can effectively improve the generalization ability.