Local Optimization Algorithms for Maximum Planar Subgraph

Consider the NP-hard problem of, given a simple graph G, to find a planar subgraph of G with the maximum number of edges. This is called the Maximum Planar Subgraph problem and the best known approximation is 4/9 and is obtained by sophisticated Graphic Matroid Parity algorithms. Here we show that applying a local optimization phase to the output of this known algorithm improves this approximation ratio by a small ϵ = 1/747 > 0. This is the first improvement in approximation ratio in more than a quarter century. The analysis relies on a more refined extremal bound on the Lovász cactus number in planar graphs, compared to the earlier (tight) bound of [5, 8]. A second local optimization algorithm achieves a tight ratio of 5/12 for Maximum Planar Subgraph without using Graphic Matroid Parity. We also show that applying a greedy algorithm before this second optimization algorithm improves its ratio to at least 91/216 < 4/9. The motivation for not using Graphic Matroid Parity is that it requires sophisticated algorithms that are not considered practical by previous work. The best previously published [7] approximation ratio without Graphic Matroid Parity is 13/33 < 5/12.