Adaptive simulated annealing (ASA): Lessons learned

TL;DR

This paper discusses lessons learned from the development and application of the adaptive simulated annealing (ASA) algorithm, highlighting its efficiency and practical insights gained through user feedback.

Contribution

It provides practical lessons and insights from real-world use of ASA, emphasizing its efficiency and development process.

Findings

ASA samples parameter space more efficiently than previous algorithms

User feedback has guided improvements in ASA code

Lessons learned are applicable to other simulated annealing algorithms

Abstract

Adaptive simulated annealing (ASA) is a global optimization algorithm based on an associated proof that the parameter space can be sampled much more efficiently than by using other previous simulated annealing algorithms. The author's ASA code has been publicly available for over two years. During this time the author has volunteered to help people via e-mail, and the feedback obtained has been used to further develop the code. Some lessons learned, in particular some which are relevant to other simulated annealing algorithms, are described.