Combined kernel KELM prediction model based on quantum particle swarm optimization
Yibo Li, Liu Fang · 2019 3rd International Conference on Electronic Information Technology and Computer Engineering (EITCE) · 2019
Extreme learning machine is a new type single hidden layer feedforward neural network, which overcome the weakness of traditional error hack propagation method, and increase the training speed of algorithm as well as decrease the adjust time of parameter. Kernel idea was used to extreme learning machine by kernel extreme learning machine to instead random map. Kernel extreme learning machine has more looser restrictions and fewer parameters as well as more simple computational complexity and better generalization, in response to the insufficient fitting ability of kernel extreme learning machine, a kernel extreme learning machine based on multi-scale wavelet kernel and polynomial combination was presented in this paper. The presented algorithm optimize combined kernel parameters, weights, and penalty factors, and test on the UCl dataset. The experiment shows that the combined nuclear learning machine owns higher prediction accuracy and other advantages compared with SVM and single core learning machine.