The shortest tour I saw in my performance testing was 5761718 when recreating the same ruin using one greedy step. The super-cheap cluster reached 5761950 after 10000 trials.With this commit option "-F" got added, forcing ruined cities to be among those from initial radial ruin:
https://github.com/Hermann-SW/RR/commit ... 351de28624
Initial ruin of 1,000 cities around city 68605 resulted in tour length 5,763,082.
I did run for 100,000 mutations and it finished at length 5,758,491 after 1:27h (optimal length is 5,757,191).
Then I started with 1 million mutations last night, and it took 14:25h and ended with length 5,758,019.
Localizing the ruin and recreate seems like a good idea. What do you think about decreasing the size of the additional mutations at the end of a run?
Statistics: Posted by ejolson — Wed Sep 17, 2025 8:37 pm