Research map / Image generation / GANs
BourGAN: Generative Networks with Metric Embeddings
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 |
|---|---|---|---|
| Use metric-embedding-based Gaussian-mixture latent sampling and the distance-preservation objective instead of standard GAN training | GAN · 2D Ring | Wasserstein-1: 1.412 -> 0.687 | Helped |
| Use metric-embedding-based Gaussian-mixture latent sampling and the distance-preservation objective instead of standard GAN training | GAN · 2D Grid | # modes: 24.3 -> 25.0 | Helped |
| Use metric-embedding-based Gaussian-mixture latent sampling and the distance-preservation objective instead of standard GAN training | GAN · 2D Circle | Wasserstein-1: 0.263 -> 0.081 | Helped |
| Use classifier distance rather than ℓ2 distance | BourGAN · MNIST | KL divergence: 0.104 -> 0.012 | Helped |
| Remove the pairwise distance-preservation term by setting β to zero (ablation) | BourGAN · Synthetic convergence example | training convergence | Hurt |
| Add generator pre-training using the correspondence between data items and embedded vectors (ablation) | BourGAN · MNIST | training convergence | Helped |
| Vary Gaussian-mixture standard deviation σ within [0.2, 0.6] (ablation) | BourGAN · Synthetic and real-data experiments | mode coverage: [0.2, 0.6] | No clear effect |
Closest papers by meaning
- Mixture Density Generative Adversarial Networks (CVPR 2019 · similarity 0.90)
- Unveiling the Latent Space Geometry of Push-Forward Generative Models (ICML 2023 · similarity 0.89)
- A Geometric Analysis of Deep Generative Image Models and Its Applications (ICLR 2021 · similarity 0.89)
- Improving GAN with Neighbors Embedding and Gradient Matching (AAAI 2019 · similarity 0.89)
- Geometry Score: A Method For Comparing Generative Adversarial Networks (ICML 2018 · similarity 0.89)
- Structure-preserving GANs (ICML 2022 · 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.