Research map / Probabilistic and scientific modeling / Variational and generative sampling
Cooperative Training of Descriptor and Generator Networks
Jianwen Xie, Yang Lu, Ruiqi Gao, Song Chun Zhu, Ying Nian Wu
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 |
|---|---|---|---|
| train descriptor and generator cooperatively instead of generator alone (Algorithm G) (ablation) | CoopNets · ImageNet-1k scene categories (n=1100) | Inception score: 2.61 -> 3.79 | Improved |
| train cooperatively instead of descriptor alone (persistent CD) (ablation) | CoopNets · ImageNet-1k scene categories (n=1100) | Inception score: 1.51 -> 3.79 | Improved |
| CoopNets instead of DCGAN with few training images | CoopNets · ImageNet-1k scene categories (n=50) | Inception score: 2.26 -> 2.66 | Improved |
| use descriptor with Langevin revision instead of generator output (ablation) | CoopNets · LSUN bedrooms | FID: 64.30 -> 35.42 | Improved |
| use descriptor with Langevin revision instead of generator output (ablation) | CoopNets · CelebA | FID: 16.98 -> 16.65 | Improved |
| use descriptor with Langevin revision instead of generator output (ablation) | CoopNets · Cifar-10 | FID: 35.25 -> 33.61 | Improved |
| CoopNets generator instead of DCGAN | CoopNets · CelebA | FID: 21.40 -> 16.98 | Improved |
| CoopNets generator instead of GAN | CoopNets · MNIST (Parzen window) | log-likelihood: 225 -> 226 | Improved |
| CoopNets instead of DCGAN for inpainting | CoopNets · CelebA pattern completion (M30 mask) | recovery error: 0.211 -> 0.115 | Improved |
| CoopNets instead of MRF-l1 inpainting at large mask | CoopNets · CelebA pattern completion (M50 mask) | recovery error: 0.178 -> 0.136 | Improved |
| CoopNets versus DCGAN in human ranking | CoopNets · LSUN bedrooms | average human ranking: 1.913 -> 1.700 | Improved |
Closest papers by meaning
- A Tale of Two Flows: Cooperative Learning of Langevin Flow and Normalizing Flow Toward Energy-Based Model (ICLR 2022 · similarity 0.89)
- Learning Generative ConvNets via Multi-Grid Modeling and Sampling (CVPR 2018 · similarity 0.88)
- Representation Alignment for Generation: Training Diffusion Transformers Is Easier Than You Think (ICLR 2025 · similarity 0.86)
- Boosting Generative Image Modeling via Joint Image-Feature Synthesis (NeurIPS 2025 · similarity 0.86)
- U-Nets as Belief Propagation: Efficient Classification, Denoising, and Diffusion in Generative Hierarchical Models (ICLR 2025 · similarity 0.85)
- Improved Distribution Matching Distillation for Fast Image Synthesis (NeurIPS 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.
