Kernel method for pattern analysis

Author: (UK) Xiao-Tai Li / (USA) Christopher N. Smith / Zhao Lingling, etc.
Publisher:
Publish Date: 2006-01-01
Features: This book provides a detailed introduction to the basic concepts and applications of kernel-based pattern analysis. The main content includes: fundamental theoretical foundations, several kernel-based algorithms, ranging from simple to more complex systems, such as kernel partial order least squares, canonical correlation analysis, support vector machines, and principal component analysis, etc. It also describes various kernel functions, from basic examples to advanced recursive kernels, kernel functions derived from generative models (e.g., HMMs), to string matching kernels based on dynamic programming, as well as special kernels for processing text documents, etc. This book is suitable for all students, teachers, and researchers engaged in pattern recognition, machine learning, neural networks, and their applications. Pattern analysis is the process of finding general relationships from a batch of data. It has gradually become the core of many disciplines, from neural networks to what is known as syntactic pattern recognition, from statistical pattern recognition to machine learning and data mining. The applications of pattern analysis cover a wide range of fields, from complex bioinformatics to relatively simple document retrieval, etc. The kernel methods described in this book provide a powerful and unified framework for all these disciplines, driving the development of various algorithms that can be applied to data of various general forms (such as strings, vectors, text, etc.) and can be used to find various types of general relationships (such as sorting, classification, regression, and clustering, etc.). This book has two main purposes. First, it provides professionals with an inclusive toolbox containing various easily implementable algorithms, kernel functions, and solutions. Many algorithms are provided with MATLAB code, making them applicable to pattern analysis tasks in many fields. Second, it serves as a convenient introductory guide for students and researchers, helping them understand the rapidly evolving field of kernel-based pattern analysis. The book illustrates how to manually write an algorithm or kernel function for a new specific application, along with preliminary plans and mathematical tools needed to accomplish this task. The book is divided into three parts. Part 1 introduces the basic concepts of this field, not only providing an expanded introductory example but also elaborating on the main theoretical foundations of this method. Part 2 includes several kernel-based algorithms, ranging from simple to more complex systems, such as kernel partial order least squares, canonical correlation analysis, support vector machines, and principal component analysis, etc. Part 3 describes various kernel functions, from basic examples to advanced recursive kernels, from kernel functions derived from generative models (e.g., HMMs) to string matching kernels based on dynamic programming, as well as special kernels for processing text documents, etc. ... This book provides a detailed introduction to the basic concepts and applications of kernel-based pattern analysis. The main content includes: fundamental theoretical foundations, several kernel-based algorithms, ranging from simple to more complex systems, such as kernel partial order least squares, canonical correlation analysis, support vector machines, and principal component analysis, etc. It also describes various kernel functions, from basic examples to advanced recursive kernels, from kernel functions derived from generative models (e.g., HMMs) to string matching kernels based on dynamic programming, as well as special kernels for processing text documents, etc. This book is suitable for all students, teachers, and researchers engaged in pattern recognition, machine learning, neural networks, and their applications.