Research map / Optimization / Implicit bias and generalization
Alternating Back-Propagation for Generator Network
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 ABP instead of PCA | CelebA image reconstruction · CelebA test images, d=20 | reconstruction error: .1038 -> .0810 | Helped |
| Use ABP instead of PCA | CelebA image reconstruction · CelebA test images, d=60 | reconstruction error: .0820 -> .0617 | Helped |
| Use ABP instead of PCA | CelebA image reconstruction · CelebA test images, d=100 | reconstruction error: .0722 -> .0549 | Helped |
| Use ABP instead of PCA | CelebA image reconstruction · CelebA test images, d=200 | reconstruction error: .0621 -> .0523 | Helped |
| Train ABP using observed pixels only | CelebA occluded-image recovery · P.5, P.7, P.9, M20, M30 | recovery error: .0571; .0662; .0771; .0773; .1035 | Mixed |
| Train ABP from random projections | CelebA compressively sensed-image recovery · CelebA, d=20 | recovery error: .0795 | Helped |
| Train ABP from random projections | CelebA compressively sensed-image recovery · CelebA, d=60 | recovery error: .0617 | Helped |
| Train ABP from random projections | CelebA compressively sensed-image recovery · CelebA, d=100 | recovery error: .0625 | Helped |
| Increase latent dimension from d=20 to d=60 (ablation) | ABP · CelebA compressively sensed images | recovery error: .0795 -> .0617 | Helped |
| Increase latent dimension from d=60 to d=100 (ablation) | ABP · CelebA compressively sensed images | recovery error: .0617 -> .0625 | Hurt |
Closest papers by meaning
- Learning Dynamic Generator Model by Alternating Back-Propagation through Time (AAAI 2019 · similarity 0.86)
- Alternating Gradient Flows: A Theory of Feature Learning in Two-layer Neural Networks (NeurIPS 2025 · similarity 0.81)
- Noise Consistency Training: A Native Approach for One-step Generator in Learning Additional Controls (NeurIPS 2025 · similarity 0.81)
- MCGAN: Enhancing GAN Training with Regression-Based Generator Loss (AAAI 2025 · similarity 0.81)
- Neural Residual Diffusion Models for Deep Scalable Vision Generation (NeurIPS 2024 · similarity 0.81)
- Network Inversion of Convolutional Neural Nets (Student Abstract) (AAAI 2025 · similarity 0.80)
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.