Skip to content

Conversation

jcoupey
Copy link
Collaborator

@jcoupey jcoupey commented Sep 14, 2023

Issue

Fixes #987

Tasks

  • Store find_top_3_insertions results in SolutionState
  • Test using various gain upper bounds for early aborts
  • Update CHANGELOG.md
  • review

@jcoupey
Copy link
Collaborator Author

jcoupey commented Sep 15, 2023

Results so far are not overwhelming but still interesting. There are small differences in the solutions since baf8109 cuts some branches assuming the triangular inequality holds on costs. Computing times reported below with a dev build.

Computing times (ms) on the usual CVRP benchmarks across x values:

x master baf8109 Delta (%)
0 349 324 -7.2
1 1267 1128 -11.0
2 2879 2672 -7.2
3 5082 4830 -5.0
4 9803 9305 -5.1
5 15587 14916 -4.3

Computing times (ms) on the usual VRPTW benchmarks using x=4:

  master baf8109 Delta (%)
Solomon 438 428 -2.3
Homberger 200 1888 1782 -5.6
Homberger 400 8804 8465 -3.9
Homberger 600 20964 20315 -3.1
Homberger 800 40057 39055 -2.5
Homberger 1000 67054 66555 -0.7

@jcoupey jcoupey merged commit 95b30e8 into master Sep 19, 2023
@jcoupey jcoupey deleted the enhancement/swap-start branch September 19, 2023 10:23
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

Reduce SWAP* computational burden
1 participant