Research map / Probabilistic and scientific modeling / Variational and generative sampling
Wasserstein Training of Restricted Boltzmann Machines
Grégoire Montavon, Klaus-Robert Müller, Marco Cuturi
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Train RBM with smoothed Wasserstein distance objective (with KL pretraining) instead of KL/MLE | Restricted Boltzmann machine · MNIST-small, PLANTS, MNIST-code: Wasserstein distance evaluation | smooth Wasserstein distance W_gamma to test data | Improved |
| Train RBM with Wasserstein objective instead of standard KL training | Restricted Boltzmann machine · Data completion on MNIST-small, PLANTS, MNIST-code | expected Hamming error | Improved |
| Train RBM with Wasserstein objective instead of standard KL training | Restricted Boltzmann machine · Data denoising on MNIST-small, PLANTS, MNIST-code | expected Hamming error | Improved |
| Omit KL pretraining before Wasserstein training (ablation) | RBM-W · MNIST-small, PLANTS, MNIST-code | Wasserstein distance | Hurt |
| Use a smaller KL weighting lambda (too little KL regularization) (ablation) | RBM-W · MNIST-small, PLANTS, MNIST-code | Wasserstein distance | Hurt |
| Increase Wasserstein smoothing parameter gamma (ablation) | RBM-W · MNIST-small, PLANTS, MNIST-code | degree of shrinkage of learned distribution | No clear effect |
Closest papers by meaning
- Asymptotic Guarantees for Learning Generative Models with the Sliced-Wasserstein Distance (NeurIPS 2019 · similarity 0.88)
- A Gradual, Semi-Discrete Approach to Generative Network Training via Explicit Wasserstein Minimization (ICML 2019 · similarity 0.88)
- Learning with Restricted Boltzmann Machines: Asymptotics of AMP and GD in High Dimensions (NeurIPS 2025 · similarity 0.86)
- Statistical and Computational Guarantees of Kernel Max-Sliced Wasserstein Distances (ICML 2025 · similarity 0.86)
- Sliced Wasserstein Estimation with Control Variates (ICLR 2024 · similarity 0.85)
- Flowing Datasets with Wasserstein over Wasserstein Gradient Flows (ICML 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
