evanfields / travelingsalesmanheuristics.jl Goto Github PK

The docs need a once-over for recent Julia versions.

• ]add instead of Pkg.add, perhaps
• Don't refer to Nullable in the simulated annealing example
• Make sure the example code loads the correct stdlibs, including packages I personally have in my startup file :)
• etc

Question, why not use the lowerbound in the sequence of heuristics to stop when a solution would be found with a cost=lowerbound?

Seems like Travis often causes problems and the ecosystem has moved to GH actions. We should follow.

This issue is used to trigger TagBot; feel free to unsubscribe.

If you haven't already, you should update your TagBot.yml to include issue comment triggers.
Please see this post on Discourse for instructions and more details.

If you'd like for me to do this for you, comment TagBot fix on this issue.
I'll open a PR within a few hours, please be patient!

AbstractArrays may have arbitrary indices; we should make sure we're not using @inbounds out of bounds and that we're not implicitly assuming user-provided arrays are all one-based indexing.

I've been (recreationally) learning about some TSP techniques and implemented an exact (slow) method here, with a specialized implementation for the symmetric case: https://github.com/ericphanson/TSP_MIP.jl. I thought it could be useful for your package for generating test cases for benchmarking heuristic performance, and also the exact solver could benefit from using the heuristic solutions as a warmstart (but that's not implemented yet in the GPLK wrapper, tracked here I believe: jump-dev/GLPK.jl#82). I actually changed the API a bit to match yours so it should be easy to use together; I saw the discourse post just now and could use such an API for the exact solver too (I decided to open an issue to not derail that thread). I'm also happy to contribute it to your package in a pull request if that's something you'd be interested in.

Hey I'm wondering whether you intend to further develope this package and if so whether you think it's more for symmetric TSPs or asymmetric ones. If you care about only symmetric ones I would like to make a PR to improve the speed of 2-opt and maybe implement the general k-opt.
Otherwise it might be reasonable to define it somehow whether a problem is symmetric or not and then handle it with different functions.