Exact Optimisation Using Simulated Annealing, Adaptive Temperature Reduction, and Sequential Monte Carlo


Speaker


Abstract

Simulated annealing is a well-established approach to optimisation that is robust for irregular objective functions. Recently it has been improved using sequential Monte Carlo. This paper presents further improvements that yield the global optimum up to machine precision. Performance is illustrated using a standard set of six test problems. Compared with existing simulated annealing approaches that for some problems cannot even approximate the global optimum the improved algorithm yields exact maximums with fewer function evaluations. The keys to this performance are algorithmic reduction of temperature and specification of Metropolis random walks, both of which utilise information inherent in the massively parallel stream of particles provided by sequential Monte Carlo. These attributes of the algorithm drastically reduce the need for problem-specific tinkering.

This event is organised by the Econometric Institute.
Twitter: @MetricsSeminars