Maximum weighted matching #23
Conversation
Codecov Report
@@ Coverage Diff @@
## master #23 +/- ##
=====================================
Coverage 100% 100%
=====================================
Files 13 14 +1
Lines 210 234 +24
=====================================
+ Hits 210 234 +24
Continue to review full report at Codecov.
|
src/matching/matching.jl
Outdated
| @@ -1,7 +1,7 @@ | |||
| __precompile__(true) | |||
| module Matching | |||
| using LightGraphs | |||
| export MatchingResult, maximum_weight_maximal_matching, minimum_weight_perfect_matching | |||
| export MatchingResult, max_weight_matching, maximum_weight_maximal_matching, minimum_weight_perfect_matching | |||
There was a problem hiding this comment.
max_weight_matching->maximum_weight_matching
for coerence with the other functions
src/matching/max_weight_matching.jl
Outdated
|
|
||
| Given a graph `g` and an edgemap `w` containing weights associated to edges, | ||
| returns a matching with the maximum total weight. | ||
| `w` is a dictionnary that maps edges i => j (with i <= j) to weights. |
|
"- If the graph is not bipartite, then the computation time may grow exponentially." I do not see integer programming part in case of fractionary solution, why do you say that? I see three options here:
|
|
also, coverage should be increased |
|
Sorry I wasn't clear in my comments, the integer programming is taken care of by the solver (by using the Int keyword in @variable, line 45). |
|
|
||
| Given a graph `g` and an edgemap `w` containing weights associated to edges, | ||
| returns a matching with the maximum total weight. | ||
| `w` is a dictionnary that maps edges i => j to weights. |
| The package JuMP.jl and one of its supported solvers is required. | ||
|
|
||
| Returns MatchingResult containing: | ||
| - a solve status (indicating whether the problem was solved to optimality) |
There was a problem hiding this comment.
MatchingResult doesn't contain solve
There was a problem hiding this comment.
ok, I saw the TODO comment
| Returns MatchingResult containing: | ||
| - a solve status (indicating whether the problem was solved to optimality) | ||
| - the optimal cost | ||
| - a list of each node's match. |
There was a problem hiding this comment.
-> a list of each vertex's match (or -1 for unmatched vertices)
test/matching/runtests.jl
Outdated
| w[Edge(1,3)] = 1 | ||
| w[Edge(1,4)] = 3 | ||
| w[Edge(2,4)] = 1 | ||
| match = Matching.maximum_weight_matching(g,w) |
There was a problem hiding this comment.
can you remove Matching. from here and from the other lines in this file (also a check to see that everything is correctly exported)
Ok, thanks for the clarification, my knowledge of JuMP is very limited. This looks good, i can be merged once the remaining comments are addressed |
|
this case should be tested if setdiff(edge_list, keys(w)) != [] # If some edges do not have a key in w.
error("Some edge weights are missing, check that keys i => j in w satisfy i <= j")
endwith a @test_throws ErrorException .... |
|
Just a quick follow-up: can this PR be merged? |
Hi,
This is a formulation for maximum weighted matching, works for both bipartite and non-bipartite graphs.
In the case of bipartite graphs, the solution of the LP is automatically integral (i.e. no need to enforce integrality constraints).
In the case of non-bipartite graphs, it requires an MIP solver (e.g. GLPK) and may therefore be much slower to solve to optimality.
Disclaimer: This is the first time I do a pull request, so I hope I'm not missing an important step.