深度强化学习结合图注意力模型求解TSP问题
王扬, 陈智斌, 杨笑笑, 吴兆蕊

Deep reinforcement learning combined with graph attention model to solve TSP
Yang Wang, Zhibin Chen, Xiaoxiao Yang, Zhaorui Wu
表4 本文模型在训练和推理阶段的时间花费
Table 4 Time cost for training and reasoning by our model
阶段TSP20TSP50TSP100
训练模型3 h24 h136 h
推理(single trajec)≪1 s5 s8 s
推理(no augment)≪1s10 s55 s
推理(8×augment)17 s1 min7 min
推理(4×augment)9 s42 s3 min