Minimax estimation of discontinuous optimal transport maps: The semi-discrete case

Authors: Aram-Alexandre Pooladian, Vincent Divol, Jonathan Niles-Weed

ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We confirm these results in numerical experiments, and provide experiments for other settings, not covered by our theory, which indicate that the entropic estimator is a promising methodology for other discontinuous transport map estimation problems.
Researcher Affiliation Academia 1Center for Data Science, New York University, NY, USA 2CEREMADE, Universit e Paris Dauphine-PSL, Paris, France 3Courant Institute of Mathematical Sciences, New York University, NY, USA.
Pseudocode No The paper does not contain any pseudocode or clearly labeled algorithm blocks.
Open Source Code No The paper does not provide an explicit statement or link for the open-sourcing of the code for its methodology.
Open Datasets No The paper uses synthetic data generated by the authors for their experiments, as stated in Section 4.3: "To create the following plots, we draw two sets of n i.i.d. points from P, (X1, . . . , Xn) and (X 1, . . . , X n), and create target points Yi = T0(X i), where T0 is known to us in advance in order to generate the data." It does not refer to a publicly available or open dataset.
Dataset Splits No The paper does not explicitly provide training/validation/test dataset splits. It describes generating synthetic data for evaluation, but without specifying partitioning ratios.
Hardware Specification No The paper does not provide any specific details about the hardware used to run its experiments (e.g., GPU/CPU models, memory).
Software Dependencies No The paper does not list specific software dependencies with version numbers (e.g., library names with versions).
Experiment Setup Yes To create the following plots, we draw two sets of n i.i.d. points from P, (X1, . . . , Xn) and (X 1, . . . , X n), and create target points Yi = T0(X i), where T0 is known to us in advance in order to generate the data. Our estimators are computed on the data (X1, . . . , Xn) and (Y1, . . . , Yn), and we evaluate the Mean-Squared error criterion MSE( ˆT) = ˆT T0 2 L2(P ) of a given map estimator ˆT using Monte Carlo integration, using 50000 newly sampled points from P. We plot the means across 10 repeated trials, accompanied by their standard deviations. First consider P = Unif([0, 1]d) and create atoms {y1, . . . , y J} by partitioning the points along the first coordinate for all j [J]: (yj)[1] = (j 1/2)/J , (yj)[2] = = (yj)[d] = 0.5 . We choose uniform qj = 1/J for j [J].