Tong Zhang

The mathematical theory of machine learning not only explains the current algorithms but can also motivate principled approaches for the future. This self-contained textbook introduces students and researchers of AI to the main mathematical techniques used to analyze machine learning algorithms, with motivations and applications. Topics covered include the analysis of supervised learning algorithms in the iid setting, the analysis of neural networks (e.g. neural tangent kernel and mean-field analysis), and the analysis of machine learning algorithms in the sequential decision setting (e.g. online learning, bandit problems, and reinforcement learning). Students will learn the basic mathematical tools used in the theoretical analysis of these machine learning problems and how to apply them to the analysis of various concrete algorithms. This textbook is perfect for readers who have some background knowledge of basic machine learning methods, but want to gain sufficient technical knowledge to understand research papers in theoretical machine learning.

This successful textbook on predictive text mining offers a unified perspective on a rapidly evolving field, integrating topics spanning the varied disciplines of data science, machine learning, databases, and computational linguistics. Serving also as a practical guide, this unique book provides helpful advice illustrated by examples and case studies. This highly anticipated second edition has been thoroughly revised and expanded with new material on deep learning, graph models, mining social media, errors and pitfalls in big data evaluation, Twitter sentiment analysis, and dependency parsing discussion. The fully updated content also features in-depth discussions on issues of document classification, information retrieval, clustering and organizing documents, information extraction, web-based data-sourcing, and prediction and evaluation. Features: includes chapter summaries and exercises; explores the application of each method; provides several case studies; contains links to free text-mining software.

This successful textbook on predictive text mining offers a unified perspective on a rapidly evolving field, integrating topics spanning the varied disciplines of data science, machine learning, databases, and computational linguistics. Serving also as a practical guide, this unique book provides helpful advice illustrated by examples and case studies. This highly anticipated second edition has been thoroughly revised and expanded with new material on deep learning, graph models, mining social media, errors and pitfalls in big data evaluation, Twitter sentiment analysis, and dependency parsing discussion. The fully updated content also features in-depth discussions on issues of document classification, information retrieval, clustering and organizing documents, information extraction, web-based data-sourcing, and prediction and evaluation. Features: includes chapter summaries and exercises; explores the application of each method; provides several case studies; contains links to free text-mining software.

One consequence of the pervasive use of computers is that most documents originate in digital form. Widespread use of the Internet makes them readily available. Text mining – the process of analyzing unstructured natural-language text – is concerned with how to extract information from these documents. Developed from the authors’ highly successful Springer reference on text mining, Fundamentals of Predictive Text Mining is an introductory textbook and guide to this rapidly evolving field. Integrating topics spanning the varied disciplines of data mining, machine learning, databases, and computational linguistics, this uniquely useful book also provides practical advice for text mining. In-depth discussions are presented on issues of document classification, information retrieval, clustering and organizing documents, information extraction, web-based data-sourcing, and prediction and evaluation. Background on data mining is beneficial, but not essential. Where advanced concepts are discussed that require mathematical maturity for a proper understanding, intuitive explanations are also provided for less advanced readers. Topics and features: presents a comprehensive, practical and easy-to-read introduction to text mining; includes chapter summaries, useful historical and bibliographic remarks, and classroom-tested exercises for each chapter; explores the application and utility of each method, as well as the optimum techniques for specific scenarios; provides several descriptive case studies that take readers from problem description to systems deployment in the real world; includes access to industrial-strength text-mining software that runs on any computer; describes methods that rely on basic statistical techniques, thus allowing for relevance to all languages (not just English); contains links to free downloadable software and other supplementary instruction material. Fundamentals of Predictive Text Mining is an essential resource for IT professionalsand managers, as well as a key text for advanced undergraduate computer science students and beginning graduate students. Dr. Sholom M. Weiss is a Research Staff Member with the IBM Predictive Modeling group, in Yorktown Heights, New York, and Professor Emeritus of Computer Science at Rutgers University. Dr. Nitin Indurkhya is Professor at the School of Computer Science and Engineering, University of New South Wales, Australia, as well as founder and president of data-mining consulting company Data-Miner Pty Ltd. Dr. Tong Zhang is Associate Professor at the Department of Statistics and Biostatistics at Rutgers University, New Jersey.

Content-Based Audio Classification and Retrieval for Audiovisual Data Parsing is an up-to-date overview of audio and video content analysis. Included is extensive treatment of audiovisual data segmentation, indexing and retrieval based on multimodal media content analysis, and content-based management of audio data. In addition to the commonly studied audio types such as speech and music, the authors have included hybrid types of sounds that contain more than one kind of audio component such as speech or environmental sound with music in the background. Emphasis is also placed on semantic-level identification and classification of environmental sounds. The authors introduce a new generic audio retrieval system on top of the audio archiving schemes. Both theoretical analysis and implementation issues are presented. The developing MPEG-7 standards are explored. Content-Based Audio Classification and Retrieval for Audiovisual Data Parsing will be especially useful to researchers and graduate level students designing and developing fully functional audiovisual systems for audio/video content parsing of multimedia streams.

Data mining is a mature technology. The prediction problem, looking for predictive patterns in data, has been widely studied. Strong me- ods are available to the practitioner. These methods process structured numerical information, where uniform measurements are taken over a sample of data. Text is often described as unstructured information. So, it would seem, text and numerical data are different, requiring different methods. Or are they? In our view, a prediction problem can be solved by the same methods, whether the data are structured - merical measurements or unstructured text. Text and documents can be transformed into measured values, such as the presence or absence of words, and the same methods that have proven successful for pred- tive data mining can be applied to text. Yet, there are key differences. Evaluation techniques must be adapted to the chronological order of publication and to alternative measures of error. Because the data are documents, more specialized analytical methods may be preferred for text. Moreover, the methods must be modi?ed to accommodate very high dimensions: tens of thousands of words and documents. Still, the central themes are similar.

Data mining is a mature technology. The prediction problem, looking for predictive patterns in data, has been widely studied. Strong me- ods are available to the practitioner. These methods process structured numerical information, where uniform measurements are taken over a sample of data. Text is often described as unstructured information. So, it would seem, text and numerical data are different, requiring different methods. Or are they? In our view, a prediction problem can be solved by the same methods, whether the data are structured - merical measurements or unstructured text. Text and documents can be transformed into measured values, such as the presence or absence of words, and the same methods that have proven successful for pred- tive data mining can be applied to text. Yet, there are key differences. Evaluation techniques must be adapted to the chronological order of publication and to alternative measures of error. Because the data are documents, more specialized analytical methods may be preferred for text. Moreover, the methods must be modi?ed to accommodate very high dimensions: tens of thousands of words and documents. Still, the central themes are similar.

Content-Based Audio Classification and Retrieval for Audiovisual Data Parsing is an up-to-date overview of audio and video content analysis. Included is extensive treatment of audiovisual data segmentation, indexing and retrieval based on multimodal media content analysis, and content-based management of audio data. In addition to the commonly studied audio types such as speech and music, the authors have included hybrid types of sounds that contain more than one kind of audio component such as speech or environmental sound with music in the background. Emphasis is also placed on semantic-level identification and classification of environmental sounds. The authors introduce a new generic audio retrieval system on top of the audio archiving schemes. Both theoretical analysis and implementation issues are presented. The developing MPEG-7 standards are explored. Content-Based Audio Classification and Retrieval for Audiovisual Data Parsing will be especially useful to researchers and graduate level students designing and developing fully functional audiovisual systems for audio/video content parsing of multimedia streams.

The Riemann problem is the most fundamental problem in the entire field of non-linear hyperbolic conservation laws. Since first posed and solved in 1860, great progress has been achieved in the one-dimensional case. However, the two-dimensional case is substantially different. Although research interest in it has lasted more than a century, it has yielded almost no analytical demonstration. It remains a great challenge for mathematicians. This volume presents work on the two-dimensional Riemann problem carried out over the last 20 years by a Chinese group. The authors explore four models: scalar conservation laws, compressible Euler equations, zero-pressure gas dynamics, and pressure-gradient equations. They use the method of generalized characteristic analysis plus numerical experiments to demonstrate the elementary field interaction patterns of shocks, rarefaction waves, and slip lines. They also discover a most interesting feature for zero-pressure gas dynamics: a new kind of elementary wave appearing in the interaction of slip lines-a weighted Dirac delta shock of the density function. The Two-Dimensional Riemann Problem in Gas Dynamics establishes the rigorous mathematical theory of delta-shocks and Mach reflection-like patterns for zero-pressure gas dynamics, clarifies the boundaries of interaction of elementary waves, demonstrates the interesting spatial interaction of slip lines, and proposes a series of open problems. With applications ranging from engineering to astrophysics, and as the first book to examine the two-dimensional Riemann problem, this volume will prove fascinating to mathematicians and hold great interest for physicists and engineers.