Expectile Regularization for Efficient Neural Optimal Transport Training
The author proposes an efficient method, ENOT, using expectile regularization to improve Neural Optimal Transport training by approximating conjugate potentials accurately and quickly.
