Towards Global Optimization with Adaptive Simulated Annealing

Abstract

The problem of global optimization is generally intractable. Of course, the lack of a guarantee that the global minimum can be found need not proscribe a search — in many practical problems, it is not unreasonable to expect to find some good local minima. Most optimization algorithms used for lens design simply seek a local minimum in the neighbourhood of some specified starting point by continuing downhill until some flat terrain is found. In essence, the tough component of this stage in the design problem becomes finding a good starting point for such a blinkered algorithm to polish up. With the continuing, remarkable advances in computer power, a more ambitious, automated search of the configuration space is becoming a realistic option.