Improved ensemble extreme learning machine regression algorithm

Abstract

Compared with other traditional neural network algorithms, the Extreme Learning Machine (ELM) has the advantages of simple structure, fast learning speed and good generalization performance. However, there are still some shortages that restrict the further development of ELM. For example, the randomly generated input weights, biases and the ill-conditioned appearance of the hidden layer design matrix all affect the generalization performance and robustness of the ELM algorithm model. In order to overcome the adverse affects of both, an improved ensemble extreme learning machine regression algorithm (ECV-ELM) is proposed in this paper. The method first generates multiple sub CV-ELM model through AdaBoost.RT method, and the selects the best set of sub-models to integrate. The ECV-ELM algorithm makes use of ensemble learning method to complement each other among sub-models, so that the generalization performance and robustness of the algorithm are better than that of the sub-model. The results of regression experiments on multiple data sets show that the ECV_ELM algorithm can effectively reduce the influence of the ill conditioned matrix, the random input weight and bias, and has good generalization performance and robustness.