Regularized Optimal Transport is Ground Cost Adversarial
Authors: François-Pierre Paty, Marco Cuturi
ICML 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | 8. Experiments, We plot in Figure 2 the mean value of c Wε T 2(µi, νi) depending on ε, In Figure 3, we visualize the effect of the regularization ε on the ground cost cε itself, We use multidimensional scaling on the adversarial cost matrix cε (with distances between points from the same measures unchanged) to recover points in R2. |
| Researcher Affiliation | Collaboration | 1CREST / ENSAE Paris, Institut Polytechnique de Paris 2Google Brain. |
| Pseudocode | Yes | Algorithm 1 Projected (sub)Gradient Ascent for Nonnegative Adversarial Cost and Algorithm 2 Randomized (Block) Coordinate Ascent for sequential SRW |
| Open Source Code | No | No statement regarding the release of source code or a link to a code repository was found. |
| Open Datasets | No | We consider 20 measures (µi)i=1,...,10, (νj)j=1,...,10 on the red-green-blue color space identified with X = [0, 1]3. Each measure is a point cloud corresponding to the colors used in a painting, divided into two types: ten portraits by Modigliani (µi, i M) and ten by Schiele (νj, j S). (No concrete access information for these specific datasets provided). |
| Dataset Splits | No | No specific dataset split information (percentages, counts, or predefined splits) for training, validation, or testing is provided. |
| Hardware Specification | No | No specific hardware details (exact GPU/CPU models, processor types, memory amounts, or detailed computer specifications) are provided for running experiments. |
| Software Dependencies | No | No specific ancillary software details (e.g., library or solver names with version numbers) are provided. |
| Experiment Setup | No | The paper mentions 'learning rate lr' and 'MAXITER' in the algorithms, and 'projected SGD' but does not provide concrete hyperparameter values or specific training configurations (e.g., specific learning rate value, number of epochs, batch size). |