Support Vector Machine Based on Orthogonal Efficient Hinging Hyperplanes Neural Network

In this paper, we present a support vector machine classifier model based on an efficient hinging hyperplanes neural network, in order to solve the classification problem. Firstly, an efficient hinging hyperplanes network is established. The hidden layer neuron connection of the network can be regarded a directed acyclic graph. Secondly, the unique neuron generation method of the network ensures the independence of piecewise linear mapping of input features. Selecting the output of the target neuron as the kernel function of the support vector machine, the support vector machine based on efficient hinging hyperplanes is constructed. Finally, the support vector machine based on efficient hinging hyperplanes is compared with other prevailing nonlinear classifiers. It can be concluded that the support vector machine based on efficient hinging hyperplanes network can complete the classification task efficiently and ensure faster classification speed and higher classification accuracy.