paper-conference

We propose a novel approach which learns straight trajectories in flow matching without iterative rectification or heuristic minibatch optimal transport approximations.

We propose a novel theoretically-justified benchmark for continuous Entropic Optimal Transport (EOT) and Schrodinger Bridge (SB) solvers.

We propose a novel theoretically-justified benchmark for continuous Entropic Optimal Transport (EOT) and Schrodinger Bridge (SB) solvers.

We propose a novel neural algorithm for the fundamental problem of computing the entropic optimal transport (EOT) plan between probability distributions which are accessible by samples.

We show that Neural Optimal Transport (NOT) algorithm with the weak quadratic cost might learn fake plans which are not optimal. We fix this by introducing kernel weak quadratic costs which provide improved theoretical guarantees.

We present a novel neural-networks-based algorithm to compute optimal transport maps and plans for strong and weak transport costs.

We construct pairs of continuous distributions with analytically-known optimal transport (OT) solutions for the Wasserstein-1 distance to test existing WGAN solvers.

We present an algorithm to approximate the Wasserstein-2 barycenters of continuous measures via a generative model and construct Ave, celeba! dataset which can be used for quantitative evaluation of barycenter algorithms.

While optimal transport cost serves as the loss for popular generative models, we demonstrate that the optimal transport map can be used as the generative model itself.

We construct pairs of continuous measures with analytically-known optimal transport (OT) solutions for the quadratic cost to test existing parametric OT solvers including those used in implicit generative modeling.