An improved regularized extreme learning machine based on symbiotic organisms search
Boyang Zhang, Lingjie Sun, Haiwen Yuan, Jianxun Lv, Zhao Ma · 2016
In this paper, a novel data classification approach is proposed based on integration of regularized extreme learning machine and Symbiotic Organisms Search (SOS). In order to simplified the description, the new method is named as Sos-RELM, which mainly contains two phases. As is known, in compared with traditional classification paths, such as SVM, LS-SVM and BP, extreme learning machine expresses its excellent ability in term of accuracy and computing time. Hence, in the first phase, we utilize regularised extreme learning machine with the goal that the output weights can be rapidly calculated. Symbiotic Organisms Search is one of new metaheuristic algorithms with various operations to update the individuals, which outperform DE, GA, and PSO. According to this effective and efficient optimization approach, in the second phase, the set of input wights, hidden biases and regularization parameter are optimized using Symbiotic Organisms Search. And the experimental results indicates that Sos-RELM attain a good comprehensive performance.