Research map / Image generation / GANs
Pareto GAN: Extending the Representational Power of GANs to Heavy-Tailed Distributions
Todd Huster, Jeremy Cohen, Zinan Lin, Kevin Chan, Charles Kamhoua, Nandi O. Leslie, Cho-Yu Jason Chiang, Vyas Sekar
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 |
|---|---|---|---|
| use GPD-tail Pareto noise rather than uniform, normal, or lognormal noise | GAN · Keystrokes | Area: 41.0 -> 21.1 (Lognormal); 59.5 -> 21.1 (Normal); 67.8 -> 21.1 (Uniform) | Improved |
| use GPD-tail Pareto noise rather than uniform, normal, or lognormal noise | GAN · Wiki Traffic | Area: 8.6 -> 4.5 (Normal); 9.5 -> 4.5 (Lognormal); 10.3 -> 4.5 (Uniform) | Improved |
| use GPD-tail Pareto noise rather than uniform, normal, or lognormal noise | GAN · LiveJournal | Area: 3.1 -> 2.0 (Lognormal); 7.8 -> 2.0 (Normal); 15.4 -> 2.0 (Uniform) | Improved |
| use GPD-tail Pareto noise rather than uniform, normal, or lognormal noise | GAN · S&P 500 | Area: 6.5 -> 4.4 (Lognormal); 7.3 -> 4.4 (Normal); 12.6 -> 4.4 (Uniform) | Improved |
| use GPD-tail Pareto noise rather than standard noise | GAN · Keystrokes | KS: 0.014 -> 0.013 (Lognormal); 0.020 -> 0.013 (Normal); 0.017 -> 0.013 (Uniform) | Improved |
| use GPD-tail Pareto noise rather than standard noise | GAN · Wiki Traffic | KS: 0.019 -> 0.017 (Lognormal); 0.023 -> 0.017 (Normal); 0.025 -> 0.017 (Uniform) | Improved |
| use GPD-tail Pareto noise rather than standard noise | GAN · LiveJournal | KS: 0.094 -> 0.105 (Uniform); 0.103 -> 0.105 (Normal); 0.111 -> 0.105 (Lognormal) | Hurt |
| use GPD-tail Pareto noise rather than standard noise | GAN · S&P 500 | KS: 0.011 -> 0.062 (Uniform); 0.019 -> 0.062 (Normal); 0.014 -> 0.062 (Lognormal) | Hurt |
| use root-Euclidean energy distance instead of standard energy distance (ablation) | Pareto GAN · 1000-dimensional synthetic manifold | Mean Area: 197 -> 28 | Improved |
| use root-Euclidean energy distance instead of standard energy distance (ablation) | Pareto GAN · 1000-dimensional synthetic manifold | Mean Log MDist: 9.8 -> 7.7 | Improved |
| increase network width | Pareto GAN · Keystrokes, Wiki Traffic, and LiveJournal | tail Area | Improved |
Closest papers by meaning
- Robust training of implicit generative models for multivariate and heavy-tailed distributions with an invariant statistical loss (JMLR 2026 · similarity 0.90)
- How Well Generative Adversarial Networks Learn Distributions (JMLR 2021 · similarity 0.89)
- Taming the Tail in Class-Conditional GANs: Knowledge Sharing via Unconditional Training at Lower Resolutions (CVPR 2024 · similarity 0.88)
- EV-GAN: Simulation of extreme events with ReLU neural networks (JMLR 2022 · similarity 0.88)
- MCGAN: Enhancing GAN Training with Regression-Based Generator Loss (AAAI 2025 · similarity 0.87)
- Revisiting Adversarial Training Under Long-Tailed Distributions (CVPR 2024 · similarity 0.86)
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.
