Information-Greedy Global Optimisation

Опубликовано 17 августа 2016, 3:26

Optimisation is about inferring the location of the optimum of a function. An information-optimal optimiser should thus aim to collapse its belief about the location of the optimum towards a point-distribution, as fast as possible. But the state of the art rarely addresses this inference problem. Instead, it usually relies on some heuristic predicting function optima, then evaluates at the maximum of the heuristic. The reason there are no truly probabilistic optimisers yet is that they are intractable in several ways. In this talk, I will present tractable approximations for each of these issues, and arrive at a flexible global optimiser for functions under Gaussian process priors, which empirically outperforms the state of the art.