Bayesian Reasoning and Machine Learning

Livro digital

Título:
Bayesian Reasoning and Machine Learning

Autor:
David Barber

Categoria:
Tecnologia > IA

Doador:
Raffaello D. N.

Sinopse:
Most machine learning books treat probability as a tool you pick up once and move past; this one opens with a matrix mapping all 28 of its chapters against five actual university courses the author has taught from this same material, from a Graphical Models course to a Time-series Short Course, and invites the reader to choose a path through the book rather than read it cover to cover. The chapters that follow build a single connected structure instead of a grab bag of algorithms. Part I covers inference in probabilistic models, from basic graph concepts through belief networks and the junction tree algorithm. Part II turns to learning in those models, including Naive Bayes, learning with hidden variables, and Bayesian model selection. Part III is where standard machine learning topics land: nearest-neighbour classification, dimension reduction, linear and Bayesian linear models, Gaussian processes, and mixture and latent-variable models. Part IV extends the same framework to dynamical systems, covering discrete- and continuous-state Markov models and switching linear dynamical systems, before Part V closes with sampling and deterministic approximate inference for the cases exact inference can't reach. Written by David Barber and maintained on his own page since 2007 with a companion BRMLtoolbox and exercise solutions for instructors, the book's real argument is structural: supervised learning, unsupervised learning, and time-series modeling are not separate subjects, they are the same Bayesian reasoning applied to different graphs.