Salo: Combining Simulated Annealing and Local Optimization for Efficient Global Optimization

Simulated annealing is an established method for global optimization. Perhaps its most salient feature is the statistical promise to deliver a globally optimal solution. In this work, we propose a technique which attempts to combine the robustness of annealing in rugged terrain with the efficiency of local optimization methods in simple search spaces. On a variety of benchmark functions, the proposed method seems to clearly outperform a parallel genetic algorithm and adaptive simulated annealing, two popular and powerful optimization techniques.