Light Schrödinger Bridge

Authors: Alexander Korotin, Nikita Gushchin, Evgeny Burnaev

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

Reproducibility Variable Result LLM Response
Research Type Experimental We demonstrate the performance of our light solver in a series of synthetic and real-data experiments (M5), including the ones with the real biological data (M5.3) considered in related works.
Researcher Affiliation Academia Alexander Korotin 1,2, Nikita Gushchin 1, Evgeny Burnaev1,2. 1Skolkovo Institute of Science and Technology, 2Artificial Intelligence Research Institute a.korotin@skoltech.ru, n.gushchin@skoltech.ru
Pseudocode No The paper describes training and inference procedures in text and with equations, but it does not include a formal pseudocode block or algorithm listing.
Open Source Code Yes The code for our solver can be found at https://github.com/ngushchin/Light SB.
Open Datasets Yes We use data from the Kaggle competition Open Problems Multimodal Single-Cell Integration : https://www.kaggle.com/competitions/open-problems-multimodal
Dataset Splits No The paper mentions 'train data' and 'test faces' for specific experiments but does not provide explicit details about training/validation/test splits, such as percentages, counts, or references to predefined validation splits, for all experiments.
Hardware Specification No The paper states that the solver runs 'on CPU' and specifies '4 CPU cores' but does not provide details on the specific CPU model or processor used.
Software Dependencies No The paper mentions 'Py Torch' and 'Adam optimiser' but does not provide specific version numbers for these or other software dependencies.
Experiment Setup Yes We use K = 500 in all the cases. For ϵ = 10^-1 and ϵ = 10^-2, we use lr = 10^-3 and for ϵ = 2 * 10^-3 we use lr = 10 and batchsize 128. We do 10^4 gradient steps.