Markov Chain Monte Carlo In Practice ... In Practice William Chesters Fugro-Jason Many AI and vision problems can arguably only be understood by posing them in probabilistic terms. And they can, allegedly, be solved using Markov Chain Monte Carlo methods. But in spite of the many books and seminars with names like "CMC In Practice", the technique has a reputation for being difficult and disappointing. In this talk I will point out the most important reasons why MCMC can fail, and discuss how to set about making it work for you.