AI Paper Atlas

Research map / Graph learning / Graph algorithms and clustering

NeuroLKH: Combining Deep Learning Model with Lin-Kernighan-Helsgaun Heuristic for Solving the Traveling Salesman Problem

Liang Xin, Wen Song, Zhiguang Cao, Jie Zhang

NeurIPS 2021 · Graph learning › Graph algorithms and clustering · approach: Neural combinatorial optimization on graphs · 44 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use SGN-predicted edge scores and node penalties to guide LKHLKH · TSP 100Obj: 7.755071 -> 7.753332Improved
Use SGN-predicted edge scores and node penalties to guide LKHLKH · TSP 5000Obj: 51.025519 -> 50.965382Improved
Use SGN edge-score candidate sets to guide LKHLKH · CVRP 500Obj: 42.1621 -> 41.7311Improved
Use SGN edge-score candidate sets to guide LKHLKH · PDP 200Obj: 13.8390 -> 13.6246Improved
Use SGN edge-score candidate sets to guide LKHLKH · CVRPTW 200Obj: 26.1757 -> 25.4000Improved
Increase sparse graph degree from 10 to 20 (ablation)NeuroLKH · TSP500percentage of optimal edges missed in the candidate set: 0.643% -> 0.209%Improved
Increase sparse graph degree from 20 to 30 (ablation)NeuroLKH · TSP500average ranking of optimal edges: 1.646 -> 1.640Improved
Train on a mixture of uniform and clustered node distributions instead of uniform-only (ablation)NeuroLKH · TSPLIB hard instancesoptimal solutions found on average: 6.13 -> 6.79Improved
Train on a mixture of uniform and clustered node distributions instead of uniform-only (ablation)NeuroLKH · TSPLIB all 72 instancesoptimal solutions found on average: 7.92 -> 8.74Improved
Increase sparse graph degree beyond 20 (ablation)NeuroLKH · TSP500average ranking of optimal edges: 1.646 -> 1.640No clear effect
Use edge scores without fine-tuning node penalties for varied node distributionsNeuroLKH · TSPLIB easy instancesoptimal solutions found: fails to solve 11 of the 48 easy instances optimallyMixed

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy