Monte-Carlo-tree-search-for-TSP
Monte-Carlo-tree-search-for-TSP copied to clipboard
Cannot train the Att-GraphConvNet on TSP-50
Hi,
Thanks for opening source the code. I am trying to reproduce the results on Att-GraphConvNet on TSP-50, however, following the code my validation prediction average length can only achieve around 6.358, which is far worse than the performance in the original paper.
I am wondering do you have any suggestions?
Thanks!