This week I used the improved code from last week to solve a 16×15 grid with good success. I compared the performance of simulated annealing versus genetic algorithms at performing this task, running each 4 times:
16x15 completion time (in seconds) Simulated Annealing: 11228, 1476, 3151, --- Genetic Algorithm: 1702.5, 3291, 2589, 3641
From this small sample size of 4, ignoring outliers like the first simulated annealing test and the last (where the algorithm failed to finish in a reasonable time and I had to terminate it early to write this blog post), the two algorithms appear to have relatively similar performance. Perhaps a higher sample size will shed more light on how the two algorithms perform relative to each other, and both the algorithms could have their parameters optimized a bit more.
shoffing's blog 