Research map / Probabilistic and scientific modeling / Variational and generative sampling
Run-Sort-ReRun: Escaping Batch Size Limitations in Sliced Wasserstein Generative Models
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Fine-tune with Run-Sort-ReRun at large effective batch size (b=16000) instead of baseline | E2GAN · CIFAR-10 | FID: 11.26 -> 4.9 | Helped |
| Fine-tune with Run-Sort-ReRun at large effective batch size (b=16000) | E2GAN · STL-10 | FID: 25.35 -> 8.6 | Helped |
| Fine-tune with Run-Sort-ReRun at batch size 8192 | StyleGAN-v2 · LSUN Church | FID: 5.93 -> 3.14 | Helped |
| Increase effective batch size from 256 to 8192 in Run-Sort-ReRun (ablation) | E2GAN · CIFAR-10 | FID: 10.02 -> 6.99 | Helped |
| Increase effective batch size from 1024 to 8192 in Run-Sort-ReRun (ablation) | E2GAN · CIFAR-10 | FID: 7.39 -> 6.99 | Helped |
| Run-Sort-ReRun with effective batch size 1024 vs baseline (ablation) | E2GAN · CIFAR-10 | FID: 11.26 -> 7.39 | Helped |
| Use Run-Sort-ReRun on Inception features from four layers concatenated rather than final layer only | E2GAN · STL-10 | visual quality of generated images (qualitative) | Helped |
| Directly optimizing Inception feature distribution (RSR) without quality-targeted objective | GAN image quality · CIFAR-10 | sample quality | No clear effect |
Closest papers by meaning
- Distributional Sliced-Wasserstein and Applications to Generative Modeling (ICLR 2021 · similarity 0.92)
- Asymptotic Guarantees for Learning Generative Models with the Sliced-Wasserstein Distance (NeurIPS 2019 · similarity 0.90)
- Amortized Projection Optimization for Sliced Wasserstein Generative Models (NeurIPS 2022 · similarity 0.90)
- Statistically Optimal Generative Modeling with Maximum Deviation from the Empirical Distribution (ICML 2024 · similarity 0.89)
- Scalable Wasserstein Gradient Flow for Generative Modeling through Unbalanced Optimal Transport (ICML 2024 · similarity 0.87)
- Sliced Wasserstein with Random-Path Projecting Directions (ICML 2024 · similarity 0.87)
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.