Machine Learning ─ A Probabilistic Perspective

Today's Web-enabled deluge of electronic data calls for automated methods of dataanalysis. Machine learning provides these, developing methods that can automatically detect patternsin data and then use the uncovered patterns to predict future data. This textbook offers acomprehensive and self-contained introduction to the field of machine learning, based on a unified,probabilistic approach. The coverage combines breadth and depth, offering necessary backgroundmaterial on such topics as probability, optimization, and linear algebra as well as discussion ofrecent developments in the field, including conditional random fields, L1 regularization, and deeplearning. The book is written in an informal, accessible style, complete with pseudo-code for themost important algorithms. All topics are copiously illustrated with color images and workedexamples drawn from such application domains as biology, text processing, computer vision, androbotics. Rather than providing a cookbook of different heuristic methods, the book stresses aprincipled model-based approach, often using the language of graphical models to specify models in aconcise and intuitive way. Almost all the models described have been implemented in a MATLABsoftware package--PMTK (probabilistic modeling toolkit)--that is freely available online. The bookis suitable for upper-level undergraduates with an introductory-level college math background andbeginning graduate students.