Habilitationsvortrag Dr. Jürgen Lerner: "Markov Chain Monte Carlo Sampling"
Wednesday, 14. December 2016
15.15 – 16.45
FB Informatik und Informationswissenschaft
Abstract: Markov chain Monte Carlo (MCMC) sampling denotes a class of algorithms to approximately sample from the joint distribution of non-independent discrete random variables when direct sampling is computationally intractable. MCMC methods are applicable to distributions P(x) for which a function proportional to the probability P can be computed efficiently but where P itself is computationally intractable, for instance, due to a normalization factor that involves summation over the whole probability space. General and locally verifiable conditions ensure that MCMC sampling converges to the desired distribution if the number of simulation steps tends to infinity.