LAUSR

Extreme learning machine (ELM) is a novel single hidden layer feed-forward network, which has become a research hotspot in various domains. Through in-depth analysis on ELM, there are four factors mainly affect its model performance, such as the input data, the input weights, the number of hidden layer nodes and the hidden layer activation function. In order to enhance the performance of ELM, an ameliorated extreme learning machine, namely AELM, is proposed based on the aforementioned four factors. The proposed method owns new way to generate input weights and bias of hidden layer and has a new-type hidden layer activation function. Simulations on many UCI benchmark regression problems have demonstrated that the AELM generally outperforms the original ELM as well as several variants of ELM. Simultaneously, the AELM is adopted to build thermal efficiency model and NOx emission model of a 330MW circulating fluidized bed boiler. The results demonstrate the AELM is a useful machine learning tool.