Improvements On Twin Support Vector Machines And Its Application

Twin Support vector machine was first proposed by Professor Jayadeva in2007.This algorithm mainly aim at generating two nonparallel planes such that each plane is closer to one of the two classes and is as far as possible from the other. As a novel support vector machine, the main difference between the classic support vector machine is that we solve a pair of quadratic programming problems (QPPs) instead of a single QPP. Meanwhile, there are two discriminate curves in TWSVM. The training speed of TWSVM is four times faster than that of SVM.TWSVM has made great progress in recent years. Currently. TWSVM has been successfully used in Pattern Recognition, Data Classification, Function Fitting and so on. The author delved into TWSVM, and further extended to least squares twin support vector machine (LSTSVM). Finally, we proposed a novel mathematical model of the LSTSVM and its incremental learning method. Meanwhile, we applied this novel mathematical model to signal classification on cognitive radio, and got positive effect on experiments.The main content of this article is as follows:(1) We study TWSVM and LSTSVM.We study two basic principles for TWSVM and LSTSVM, and establish these mathematical models. Compared with the tradition support vector machine, both approaches have good generalization on some UCI experiments.(2) We proposed an improved classification algorithm (ILSTSVM) based on LSTSVM and its incremental learning version (IILSTSVM).According to the mathematical model of LSTSVM, we proposed an improved classification algorithm (ILSTSVM), and proved that this algorithm has great recognition rate by some comparison experiments between LSSVM and LSTSVM. Furthermore, we proposed a novel incremental learning algorithm of ILSTSVM. Compared with ISVM, we can prove that this algorithm has excellent performance in the low dimensional space.(3) We apply ILSTSVM and IILSTSVM to signal identification of Cognitive Radio SystemSpectrum sensing is one of key technology of cognitive radio system. And signal identification is important content of spectrum sensing. We apply improved least squares twin support vector machine and its incremental learning version to signal identification of cognitive radio system. And we have good performance on recognition rate.