基于差分进化的在线极限学习机

Online sequential extreme learning machine (OS-ELM) is an online sequential learning algorithm, which can learn data one-by-one or chunk-by-chunk. It has been shown that OS-ELM runs much faster than other popular sequential learning algorithms. In order to improve the generalization performance of OS-ELM,we propose online sequential extreme learning machine based on the differential evolution(DEOS-ELM): firstly encode the connection weights and thresholds as the fitness index of differential evolution to get the optimal structure using differential evolution algorithm global optimization ability, then predict the new data with the OS-ELM network optimized. The results show that the DEOS-ELM has better generalization performance and stability than the original OS-ELM.