Efficient hierarchical MCMC for policy search

Strens, Malcolm
Efficient hierarchical MCMC for policy search
International Conference on Machine Learning, 2004 (pdf - 318KB)

Abstract: Many inference and optimization tasks in machine learning can be solved by sampling approaches such as Markov Chain Monte Carlo (MCMC) and simulated annealing. These methods can be slow if a single target density query requires many runs of a simulation (or a complete sweep of a training dataset). We introduce a hierarchy of MCMC samplers that allow most steps to be taken in the solution space using only a small sample of simulation runs (or training examples). This is shown to accelerate learning in a policy search optimization task.