Pathwise optimal transport bounds between a one-dimensional diffusion and its Euler scheme (Q2454403)

Let \(X\) be a solution of a one-dimensional stochastic differential equation and let \(\bar{X}\) be its Euler discretization with \(N\) steps. Let \(\mathcal{W}(\mathcal{L}(X),\mathcal{L}(\bar{X}))\) be the Wasserstein distance between the pathwise laws \(\mathcal{L}(X)\) and \(\mathcal{L}(\bar{X}).\) The authors prove that \(\mathcal{W}(\mathcal{L}(X),\mathcal{L}(\bar{X}))\) behaves like \(\mathcal{O}(N^{-2/3+\varepsilon})\), where \(\varepsilon\) is an arbitrary positive constant. This rate is intermediate between the strong error estimation \(\mathcal{O}(N^{-1/2})\) and the weak error estimation \(\mathcal{O}(N^{-1}).\) They also prove that the supremum over \(t\in [0,T]\) of the Wasserstein distance between the laws \(\mathcal{L}(X_{t})\) and \(\mathcal{L}(\bar{X}_{t})\) is \(\mathcal{O}(\sqrt{\log N}N^{-1})\).