Research map / Probabilistic and scientific modeling / Variational and generative sampling
Can Push-forward Generative Models Fit Multimodal Distributions?
Antoine Salmona, Valentin De Bortoli, Julie Delon, Agnès Desolneux
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 a direct Lipschitz push-forward with a constrained Lipschitz constant to model separated modes | VAEs and GANs · Balanced two-Gaussian mixtures | distribution fit | Hurt |
| Use a VAE decoder rather than a GAN generator | VAE and GAN · Univariate balanced two-Gaussian mixture | interpolation between modes | Mixed |
| Use a score-based generative model instead of a direct push-forward model | VAE and GAN · Univariate balanced two-Gaussian mixture | generation of separated modes | Improved |
| Apply spectral normalization to the generator (ablation) | GAN · Univariate multimodal mixture | mode coverage | Hurt |
| Increase the targeted generator Lipschitz constant with gradient penalty (ablation) | GAN · Univariate multimodal mixture | mode coverage and mode collapse | Mixed |
| Increase the depth of the decoder (ablation) | VAE · Univariate balanced two-Gaussian mixture | decoder Lipschitz constant | Improved |
| Increase model size by adding depth (ablation) | GAN · Univariate balanced two-Gaussian mixture | training stability | Hurt |
| Add additive ResNet skip connections (ablation) | GAN · Univariate balanced two-Gaussian mixture | training stability | Improved |
| Use more expressive ResNet or DenseNet decoder architectures (ablation) | VAE and GAN · Univariate balanced two-Gaussian mixture | generator Lipschitz constant | No clear effect |
| Use a score-based model instead of a direct push-forward model | VAE and GAN · MNIST-derived Gaussian mixture | interpolation between modes | Improved |
| Use a score-based model instead of a direct push-forward model | VAE and GAN · MNIST 3-and-7 subset | recovery of distribution structure | Mixed |
Closest papers by meaning
- Generative Modeling via Drifting (NeurIPS 2026 · similarity 0.88)
- Statistically Optimal Generative Modeling with Maximum Deviation from the Empirical Distribution (ICML 2024 · similarity 0.86)
- Improved Diffusion-based Generative Model with Better Adversarial Robustness (ICLR 2025 · similarity 0.85)
- Neural Diffusion Models (ICML 2024 · similarity 0.85)
- Unified Generation, Reconstruction, and Representation: Generalized Diffusion with Adaptive Latent Encoding-Decoding (ICML 2024 · similarity 0.85)
- Towards Theoretical Understandings of Self-Consuming Generative Models (ICML 2024 · similarity 0.85)
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.
