MCMCsurvey - Machine Learning, 50, 543, 2003 c 2003 Kluwer...

Info iconThis preview shows pages 1–3. Sign up to view the full content.

View Full Document Right Arrow Icon
Machine Learning, 50, 5–43, 2003 c ° 2003 Kluwer Academic Publishers. Manufactured in The Netherlands. An Introduction to MCMC for Machine Learning CHRISTOPHE ANDRIEU C.Andrieu@bristol.ac.uk Department of Mathematics, Statistics Group, University of Bristol, University Walk, Bristol BS8 1TW, UK NANDO DE FREITAS nando@cs.ubc.ca Department of Computer Science, University of British Columbia, 2366 Main Mall, Vancouver, BC V6T 1Z4, Canada ARNAUD DOUCET doucet@ee.mu.oz.au Department of Electrical and Electronic Engineering, University of Melbourne, Parkville, Victoria 3052, Australia MICHAEL I. JORDAN jordan@cs.berkeley.edu Departments of Computer Science and Statistics, University of California at Berkeley, 387 Soda Hall, Berkeley, CA 94720-1776, USA Abstract. This purpose of this introductory paper is threefold. First, it introduces the Monte Carlo method with emphasis on probabilistic machine learning. Second, it reviews the main building blocks of modern Markov chain Monte Carlo simulation, thereby providing and introduction to the remaining papers of this special issue. Lastly, it discusses new interesting research horizons. Keywords: Markov chain Monte Carlo, MCMC, sampling, stochastic algorithms 1. Introduction A recent survey places the Metropolis algorithm among the ten algorithms that have had the greatest in±uence on the development and practice of science and engineering in the 20th century (Beichl & Sullivan, 2000). This algorithm is an instance of a large class of sampling algorithms, known as Markov chain Monte Carlo (MCMC). These algorithms have played a signi²cant role in statistics, econometrics, physics and computing science over the last two decades. There are several high-dimensional problems, such as computing the volume of a convex body in d dimensions, for which MCMC simulation is the only known general approach for providing a solution within a reasonable time (polynomial in d ) (Dyer, Frieze, & Kannan, 1991; Jerrum & Sinclair, 1996). While convalescing from an illness in 1946, Stan Ulam was playing solitaire. It, then, occurred to him to try to compute the chances that a particular solitaire laid out with 52 cards would come out successfully (Eckhard, 1987). After attempting exhaustive combinatorial calculations, he decided to go for the more practical approach of laying out several solitaires at random and then observing and counting the number of successful plays. This idea of selecting a statistical sample to approximate a hard combinatorial problem by a much simpler problem is at the heart of modern Monte Carlo simulation.
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
6 C. ANDRIEU ET AL. Stan Ulam soon realised that computers could be used in this fashion to answer ques- tions of neutron diffusion and mathematical physics. He contacted John Von Neumann, who understood the great potential of this idea. Over the next few years, Ulam and Von Neumann developed many Monte Carlo algorithms, including importance sampling and rejection sampling. Enrico Fermi in the 1930’s also used Monte Carlo in the calculation of
Background image of page 2
Image of page 3
This is the end of the preview. Sign up to access the rest of the document.

Page1 / 39

MCMCsurvey - Machine Learning, 50, 543, 2003 c 2003 Kluwer...

This preview shows document pages 1 - 3. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online