Purr is curled up on the Xeon server in anticipation of more runs. Scratchy claims the difficulty lies with Fido's favorite random numbers and more runs will not help. Shy looked shyly at the ground thinking about all the possible bugs in the C code.You really need to do more runs.
The total number of different reinsertion orderings is 1000! which is a very big number, and 1000 is nothing compared to it.
I did run with same seed 1234 as before, showing the 5940 longer local minimum:Then I increased seed by 1 and got only 5142 longer tour, outside of your boundaries on the other side:Code:
hermann@7950x:~/RR/tsp$ ./greedy -m 0 -R 68605 -S 1000 -s 1234 ../data/tsp/extra/mona-lisa100K -i ../data/tsp/extra/mona-lisa100K.opt.tour-1 init_dist() [5585us]5757191 global minimum5763151 local minimum found (0 greedy mutations; seed=1234)564 ms (only recreate)hermann@7950x:~/RR/tsp$Code:
hermann@7950x:~/RR/tsp$ ./greedy -m 0 -R 68605 -S 1000 -s 1235 ../data/tsp/extra/mona-lisa100K -i ../data/tsp/extra/mona-lisa100K.opt.tour-1 init_dist() [5578us]5757191 global minimum5762333 local minimum found (0 greedy mutations; seed=1235)564 ms (only recreate)hermann@7950x:~/RR/tsp$
On Shy's advice I made a slight change to how the random permutations are computed; however, I'd be surprised if that made a difference. From what I can tell the kittens thought the greedy algorithm was about getting more fish, so I suspect there is also a misunderstanding somewhere.
Statistics: Posted by ejolson — Sat Sep 13, 2025 2:45 am