Research map / Image generation / GANs
Flow-GAN: Combining Maximum Likelihood and Adversarial Learning in 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 |
|---|---|---|---|
| switch from MLE to adversarial training | Flow-GAN · MNIST | MODE Score: 7.42 -> 9.24 | Helped |
| switch from MLE to adversarial training | Flow-GAN · MNIST | Test NLL (nats): -3334.56 -> -1604.09 | Hurt |
| combine adversarial and maximum-likelihood objectives | Flow-GAN · MNIST | MODE Score: 7.42 -> 9.37 | Helped |
| combine adversarial and maximum-likelihood objectives | Flow-GAN · MNIST | Test NLL (nats): -3334.56 -> -3342.95 | Helped |
| switch from MLE to adversarial training | Flow-GAN · CIFAR-10 | Inception Score: 2.92 -> 5.76 | Helped |
| switch from MLE to adversarial training | Flow-GAN · CIFAR-10 | Test NLL (bits/dim): 3.54 -> 8.53 | Hurt |
| combine adversarial and maximum-likelihood objectives | Flow-GAN · CIFAR-10 | Inception Score: 2.92 -> 3.90; 5.76 -> 3.90 | Mixed |
| combine adversarial and maximum-likelihood objectives | Flow-GAN · CIFAR-10 | Test NLL (bits/dim): 3.54 -> 4.21; 8.53 -> 4.21 | Mixed |
| replace Flow-GAN ADV with a training-data-centered Gaussian mixture model | CIFAR-10 · CIFAR-10 | held-out NLL and sample quality | Helped |
| replace Flow-GAN ADV with a training-data-centered Gaussian mixture model | MNIST · MNIST | held-out NLL and sample quality | Helped |
| estimate likelihood with AIS instead of exact Flow-GAN likelihood | Flow-GAN checkpoints · MNIST | likelihood estimate fidelity and ranking: -3287.69 vs -2584.40; 26350.30 vs -2916.10; -3121.53 vs -2703.03 | Hurt |
| estimate likelihood with KDE instead of exact Flow-GAN likelihood | Flow-GAN checkpoints · MNIST | likelihood estimate fidelity and ranking: -3287.69 vs -167.10; 26350.30 vs -3.03; -3121.53 vs -205.69 | Hurt |
Closest papers by meaning
- Adversarial Flow Models (ICML 2026 · similarity 0.93)
- Adaptive Density Estimation for Generative Models (NeurIPS 2019 · similarity 0.92)
- Variational Annealing of GANs: A Langevin Perspective (ICML 2019 · similarity 0.91)
- A Large-Scale Study on Regularization and Normalization in GANs (ICML 2019 · similarity 0.90)
- MonoFlow: Rethinking Divergence GANs via the Perspective of Wasserstein Gradient Flows (ICML 2023 · similarity 0.90)
- MCGAN: Enhancing GAN Training with Regression-Based Generator Loss (AAAI 2025 · similarity 0.89)
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.