The Research On RBF-ELM Two-phase Learning Algorithm

Radial basis function-extreme learning machine(RBF-ELM) is a variant of extreme learning machine(ELM). RBF-ELM is an algorithm for training radial basis function networks. Similar to ELM, RBF-ELM also employs randomized method to initialize the centers and widths of RBF kernels, and analytically calculate the output weights of RBF networks. Randomly initialization of the two parameters will give rise to instability of RBF-ELM. Moreover, for different data sets, it is difficult to determine the number of the hidden nodes. In order to deal with the problems mentioned above, we proposed two two-phase RBF-ELM learning algorithms in this paper, namely the RBF-ELM based on multi-scale instance selection and the RBF-ELM based on core sets. The first phase of the two proposed algorithms is to elaborately select the centers of the RBF network rather than randomly initializing, the second phase is to train the RBF network with ELM. Therefore only the parameter of widths is randomly initialized. The experimental results show that the two proposed algorithms are feasible and effective and have better generalization performance compared with the RBF-ELM.