Enhancing Kernel Methods for Pattern Classification: Theories and Implementations

Kernel methods are a new family of techniques with sound theoretical grounds. They have been shown to be powerful approaches to pattern classification problems. However, many of the newly created kernel methods are far from perfect, and extensions and improvements are always required to make them even more effective. This book investigates one important class of the kernel methods, the least square support vector machines (LS-SVM), and enhances its performance extensively. In particular, the LS-SVM is enhanced in the contexts of four sub-problems related to solving the pattern classification problem. That is, model selection, feature selection, building sparse kernel classifier and kernel classifier ensemble. The LS-SVM can be regarded as a representative of many other kernel methods, and thus many ideas presented in this book can be easily extended to enhance performance of those related kernel methods. The results obtained should be useful to professionals that work on the theoretical aspects of kernel methods, or anyone else who may be considering ustilizing kernel methods for real-world pattern classification problems.

Ke Tang, Ph.D: Obtained his Ph.D degree from Nanyang Technological University, Singapore. He is currently an associate professor with the School of Computer Science and Technology, University of Science and Technology of China (USTC), Hefei, China. His research interests include machine learning, evolutionary computation and data mining.