Research map / Image generation / GANs
Adaptive Density Estimation for 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 |
|---|---|---|---|
| Add invertible feature mapping and train with both coverage and adversarial quality objectives | CIFAR-10 GAN baseline · CIFAR-10 | IS: 6.8 to 7.1 | Helped |
| Add invertible feature mapping and train with both coverage and adversarial quality objectives | CIFAR-10 GAN baseline · CIFAR-10 | FID: 31.4 to 28.0 | Helped |
| Add invertible feature mapping and train with both coverage and adversarial quality objectives | CIFAR-10 VAE baseline · CIFAR-10 test | BPD: 4.4 to 3.9 | Helped |
| Add invertible feature mapping to coverage-only training (ablation) | CIFAR-10 VAE · CIFAR-10 | IS: 2.0 to 3.0 | Helped |
| Add invertible feature mapping to coverage-only training (ablation) | CIFAR-10 VAE · CIFAR-10 | FID: 171 to 112 | Helped |
| Add residual connections to the discriminator (ablation) | CIFAR-10 AV-ADE · CIFAR-10 | FID: 28.0 to 26.0 | Helped |
| Use a wider generator with twice as many channels and residual discriminator (ablation) | CIFAR-10 AV-ADE · CIFAR-10 | FID: 28.0 to 17.2 | Helped |
| Use a wider generator with twice as many channels and residual discriminator (ablation) | CIFAR-10 AV-ADE · CIFAR-10 | IS: 7.1 to 8.2 | Helped |
| Use inverse autoregressive flow for posterior approximation and residual discriminator (ablation) | CIFAR-10 AV-ADE · CIFAR-10 test | BPD: 3.9 to 3.7 | Helped |
| Increase the feature mapping from one scale to two (ablation) | CIFAR-10 AV-ADE · CIFAR-10 | BPD: 3.77 to 3.48 | Helped |
| Increase the feature mapping from one scale to two (ablation) | CIFAR-10 AV-ADE · CIFAR-10 | FID: 20.1 to 27.7 | Hurt |
| Increase the feature mapping from two scales and two blocks to two scales and four blocks (ablation) | CIFAR-10 AV-ADE · CIFAR-10 | BPD: 3.48 to 3.46 | Helped |
Closest papers by meaning
- Flow-GAN: Combining Maximum Likelihood and Adversarial Learning in Generative Models (AAAI 2018 · similarity 0.92)
- Complexity Matters: Rethinking the Latent Space for Generative Modeling (NeurIPS 2023 · similarity 0.90)
- Unified Generation, Reconstruction, and Representation: Generalized Diffusion with Adaptive Latent Encoding-Decoding (ICML 2024 · similarity 0.90)
- Direct Discriminative Optimization: Your Likelihood-Based Visual Generative Model is Secretly a GAN Discriminator (ICML 2025 · similarity 0.90)
- Variational Annealing of GANs: A Langevin Perspective (ICML 2019 · similarity 0.90)
- Statistically Optimal Generative Modeling with Maximum Deviation from the Empirical Distribution (ICML 2024 · similarity 0.88)
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.