Fast Approximation of the Sliced-Wasserstein Distance Using Concentration of Random Projections

Authors: Kimia Nadjahi, Alain Durmus, Pierre E Jacob, Roland Badeau, Umut Simsekli

NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We validate our theoretical findings on synthetic datasets, and illustrate the proposed approximation on a generative modeling problem.
Researcher Affiliation Academia Kimia Nadjahi1 , Alain Durmus2, Pierre E. Jacob3, Roland Badeau1, Umut Sim sekli4 1: LTCI, Télécom Paris, Institut Polytechnique de Paris, France 2: Université Paris-Saclay, ENS Paris-Saclay, CNRS, Centre Borelli, F-91190 Gif-sur-Yvette, France 3: Department of Information Systems, Decision Sciences and Statistics, ESSEC Business School, Cergy, France 4: INRIA Département d Informatique de l École Normale Supérieure, PSL Research University, Paris, France
Pseudocode No The paper does not contain any structured pseudocode or algorithm blocks.
Open Source Code Yes Our empirical results can be reproduced with our open source code2. 2See https://github.com/kimiandj/fast_sw
Open Datasets Yes models on MNIST and Celeb A
Dataset Splits Yes In each setting, we generate two sets of d-dimensional samples, denoted by {x(j)}n j=1 and {y(j)}n j=1 with n = 104
Hardware Specification No The paper mentions that models were trained 'On GPU' and 'on CPU' and refers to 'GPU-accelerated implementation', but does not provide specific hardware details such as exact GPU/CPU models or processor types.
Software Dependencies No The paper states 'the models were trained using Py Torch', but does not specify its version number or any other software dependencies with their versions.
Experiment Setup Yes In each setting, we generate two sets of d-dimensional samples, denoted by {x(j)}n j=1 and {y(j)}n j=1 with n = 104