Coding Ockham's Razor

This book explores inductive inference using the minimum message length (MML) principle, a Bayesian method which is a realisation of Ockham's Razor based on information theory. Accompanied by a library of software, the book can assist an applications programmer, student or researcher in the fields of data analysis and machine learning to write computer programs based upon this principle.

MML inference has been around for 50 years and yet only one highly technical book has been written about the subject. The majority of research in the field has been backed by specialised one-off programs but this book includes a library of general MMLbased software, in Java. The Java source code is available under the GNU GPL open-source license. The software library is documented using Javadoc which produces extensive cross referenced HTML manual pages. Every probability distribution and statistical model that is described in the book is implemented and documentedin the software library. The library may contain a component that directly solves a reader's inference problem, or contain components that can be put together to solve the problem, or provide a standard interface under which a new component can be written to solve the problem.

This book will be of interest to application developers in the fields of machine learning and statistics as well as academics, postdocs, programmers and data scientists. It could also be used by third year or fourth year undergraduate or postgraduate students.

Gathers together the minimum necessary MML theory in one place" Implemented models and estimators include those for discrete, continuous and multivariate data, mixture models (clustering), regressions, classification trees, models of vectors and directions, linear regression, models of graphs (networks) All models are taken from the maths through to computer code and to use An accompanying library of software includes standard probability distributions, statistical models and estimators The e-book contains internet links to the software, documentation, and interactive (javascript) examples

Inhalt
1 Introduction.- 2 Discrete.- 3 Integers.- 4 Continuous.- 5 Function-Models.- 6 Multivariate.- 7 Mixture Models.- 8 Function-Models 2.- 9 Vectors.- 10 Linear Regression.- 11 Graphs.- 12 Bits and Pieces.- 13 An Implementation.- 14 Glossary.