Neural Combinatorial Optimization with Reinforcement Learning In PyTorch
PyTorch implementation of Neural Combinatorial Optimization with Reinforcement Learning.
I have implemented the basic RL pretraining model with greedy decoding from the paper. An implementation of the supervised learning baseline model is available here. Instead of a critic network, I got my results below on TSP from using an exponential moving average critic. The critic network is simply commented out in my code right now. From correspondence with a few others, it was determined that the exponential moving average critic significantly helped improve results.
My implementation uses a stochastic decoding policy in the pointer network, realized via PyTorch’s torch.multinomial(), during training, and beam search (not yet finished, only supports 1 beam a.k.a. greedy) for decoding when testing the model.
Currently, there is support for a sorting task