Estimating Barycenters of Distributions with Neural Optimal Transport

Authors: Alexander Kolesov, Petr Mokrov, Igor Udovichenko, Milena Gazdieva, Gudmund Pammer, Evgeny Burnaev, Alexander Korotin

ICML 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We showcase its applicability and effectiveness in illustrative scenarios and image data setups.
Researcher Affiliation Academia 1Skolkovo Institute of Science and Technology, Moscow, Russia 2Department of Mathematics, ETH Z urich, Z urich, Switzerland 3Artificial Intelligence Research Institute, Moscow, Russia.
Pseudocode Yes Algorithm 1 OT Barycenter via Neural Optimal Transport
Open Source Code Yes Our source code is available at https://github. com/justkolesov/NOTBarycenters.
Open Datasets Yes We use images from the MNIST dataset. ... We utilize Ave, celeba! barycenter benchmark as proposed in (Korotin et al., 2022).
Dataset Splits No The paper mentions hyperparameters like 'batch size' and '# of epochs' in Table 4, but it does not explicitly state specific train/validation/test splits with percentages or sample counts for reproduction.
Hardware Specification No The paper does not provide specific hardware details such as GPU models (e.g., 'NVIDIA A100'), CPU types, or cloud computing instance specifications used for the experiments.
Software Dependencies No The paper mentions using the 'Style GAN2-ada model' from an official repository, but it does not specify software dependencies with version numbers (e.g., 'Python 3.8', 'PyTorch 1.9') used for their implementation.
Experiment Setup Yes We aggregate the hyper-parameters of our Algorithm 1 for different experiments in Table 4.