Using NN (Neural Networks) to tackle continuous optimal transport problem is a promising approach especially for unpaired style-transfer problem. This method learns a one-to-one mapping (OT map) between the source and target data distributions but uses adversarial training (similar to GANs), which is not very stable.