Two heuristic strategies for searching optimal hyper parameters of C-SVM
暂无分享,去创建一个
Searching optimal hyper parameters of support vector machine classifier has biggish effect to the performance of the classifier. It could be found through k-fold cross-validation to penalty factor C and parameter σ in Gauss kernel function based on grid searching that the distribution of classifier precision is of isoline and multi-peak value. According to this, two heuristic search strategies are established, two-point central vertical method and multi-point barycenter method. The framework of corresponding heuristic algorithm is established based on the strategies. Simulation experiments show that the algorithm is helpful to accelerating the process searching optimal hyper parameters of SVM classifier.
[1] Chih-Jen Lin,et al. LIBSVM: A library for support vector machines , 2011, TIST.
[2] Hou Yuan-bin. Parameters selection of support vector regression by genetic algorithms , 2006 .
[3] Yang Chun-de. Anomaly detection approach based on immune algorithm and support vector machine , 2006 .
[4] Vladimir Vapnik,et al. The Nature of Statistical Learning , 1995 .