residual edge-graph attention neural network [Kun Lei, Neurocomputing, Volume 508, 2022, Pages 79-98] ▪ https://www.sciencedirect.com/science/article/abs/pii/S092523122200978X ◦ github ▪ https://github.com/Lei-Kun/DRL-and-graph-neural-network-for-routing-problems • 具体的にOR-Toolsよりどれくらい改善したか(結果の⼀部を抜粋) ◦ TSP - ノード数50 ▪ OR-Tools opt gap = 1.83% ▪ 強化学習 opt gap = 0.98% (実⾏時間4.0秒)