SVM (support vector machines) have become an increasingly popular tool for machine learning tasks involving classification, regression or novelty detection. In particular,they exhibit good generalization performance on many real issues and the approach is properly motivated theoretically. There are relatively a few free parameters to adjust and the architecture of the learning machine does not need to be found by experimentation. In this paper,survey ofthe key contents on this subject, focusing on the most well-known models based on kernel substitution, namely SVM, as well as the activated fields at present and the development tendency,is presented.