Section: New Results

Bipartite matching approximation

Participants: F. Dufossé

Bipartite matching is a classical academic problem on bipartite graphs. Many iterative heuristics need an initial approximate matching with linear computational time. We have designed two randomized highly parallelizable algorithms with linear execution time and quality guarantee. The approximation guarantees have been proved to reach respectively 63demonstrate the speed-ups and validate the applicability and efficiency of these algorithms on general bipartite graphs. Comparisons with the more efficient suboptimal linear algorithms of bipartite matching demonstrate a lower efficiently in average but a similar execution time, and validate the quality guarantee on all experiments. This work has been published in [16] .