Computing Tree Decompositions with FlowCutter: PACE 2017 Submission

TL;DR

This paper presents a novel algorithm for computing tree decompositions, which successfully solved all instances in a competitive benchmark, demonstrating its effectiveness and versatility across different problem domains.

Contribution

The paper introduces a new algorithm for tree decomposition based on FlowCutter, linking it to shortest path acceleration and parameterized complexity, and achieving top performance in PACE 2017.

Findings

Solved all benchmark instances successfully

Achieved second place in PACE 2017

Links tree decomposition to shortest path acceleration

Abstract

We describe the algorithm behind our PACE 2017 submission to the heuristic tree decomposition computation track. It was the only competitor to solve all instances and won a tight second place. The algorithm was originally developed in the context of accelerating shortest path computation on road graphs using multilevel partitions. We illustrate how this seemingly unrelated field fits into tree decomposition and parameterized complexity theory.