Need clarification about reannealing in simulannealbnd
2 views (last 30 days)
Show older comments
Hello,
I have some questions regarding 'reannealing' process of MATLAB function simulannealbnd. I am following this link. https://uk.mathworks.com/help/gads/how-simulated-annealing-works.html
1. It says that reannealing is done after a fixed number of accepted steps with the default number of steps being 100. I am trying to minimise an objective function of 5 variables. The reannealing seems to happen at the interval of every ~150 (but not exactly 150) iterations (which I presume is same as number of accepted steps). Could someone help me understand this ?
2. How does the process of reannealing converge ? Do the increase in temperatures due reannealing keep decreasing ?
3.If any of the stopping criteria is not met in between two reannealings, then how does the optimisation process ever stop ? In other words, what guarantees that the simulated annealing will converge (i.e. one of the criteria is met e.g. say objective function changes less than the FunctionTolerance) in between two reannealings ?
4. Is it possible to modify the stopping criteria so that the algorithm stops after fixed number of iterations after a best value of objective function is reached ? e.g. if objective function has best value =0.11 which does not change after ~ 2000 iterations, is it possible to make this a stopping criterion ? 6. Finally, is there a way to switch off reannealing ?
I also read the reference as given in this link but that is a general account of adaptive simulated annealing and I am not completely clear on how MATLAB adopts it, specially the reannealing part.
Thanks in advance for help --Pragya
0 Comments
Answers (0)
See Also
Categories
Find more on Simulated Annealing in Help Center and File Exchange
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!