Research map / Image generation / GANs
Variational Annealing of GANs: A Langevin Perspective
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 |
|---|---|---|---|
| dynamic oscillatory likelihood annealing (positive then negative λ) | RKL-GAN · Cifar10 | Inception score / FID: IS 6.24 to 7.05; FID 36.5 to 28.9 (λ=0 vs OA) | Helped |
| oscillatory dynamic likelihood annealing | JSD-GAN · Cifar10 | FID: FID 34.9 to 29.1 | Helped |
| oscillatory dynamic likelihood annealing | W-GAN · Cifar10 | FID: FID 31.4 to 29.0 (IS 6.86 to 6.91) | Helped |
| strong static positive likelihood regularization (large λ) (ablation) | RKL-GAN, JSD-GAN, W-GAN · Cifar10 | FID: FID worsens at λ=50 for all models (e.g. W-GAN 40.2 to 44.1 range) | Hurt |
| negative static annealing (λ<0) (ablation) | RKL-GAN · Cifar10 | FID: FID 34.5 at λ=-50 and 36.7 at λ=-10 vs 36.5 at λ=0; no consistent gain in static setting | Mixed |
| oscillatory dynamic annealing vs no annealing on W-GAN | W-GAN · Cifar10 | NLL (nats/dim): 2.96 (PMA column) to 2.37 | Helped |
| U-GAN with flow-based bridging measure (vs SVGD) for Bayesian logistic regression | U-GAN · Bayesian logistic regression (9 UCI datasets) | test NLL: heart .48 vs SVGD .45; bcancer .08 vs .08; sonar .57 vs .55; twonorm .06 vs .06 | Mixed |
| annealing of λ on unnormalized kidney distribution U-GAN (ablation) | U-GAN · Toy kidney distribution | qualitative mode coverage: no delta reported | Helped |
| removing permutation layers from the generative flow (ablation) | U-GAN · Toy restricted-flow target | qualitative fit: no delta reported | No clear effect |
Closest papers by meaning
- A Large-Scale Study on Regularization and Normalization in GANs (ICML 2019 · similarity 0.93)
- A Variational Inequality Perspective on Generative Adversarial Networks (ICLR 2019 · similarity 0.92)
- Are GANs Created Equal? A Large-Scale Study (NeurIPS 2018 · similarity 0.91)
- Bayesian GAN (NeurIPS 2017 · similarity 0.91)
- Entropic GANs meet VAEs: A Statistical Approach to Compute Sample Likelihoods in GANs (ICML 2019 · similarity 0.91)
- MCGAN: Enhancing GAN Training with Regression-Based Generator Loss (AAAI 2025 · similarity 0.90)
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.