Research map / Image generation / GANs
Forward Super-Resolution: How Can GANs Learn Hierarchical Generative Models for Real-World Distributions
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 |
|---|---|---|---|
| Layer-wise alternating training (method of moments discriminator for output layer, D4/D5 for first hidden layer, D2 super-resolution for deeper layers) with SGDA | GAN generator with forward super-resolution target · theoretical analysis (population case) | learnability of target distribution (W2 distance bound) | No clear effect |
| Initializing the first hidden layer discriminator/generator with a small bias b (b << 1) and periodically decreasing it | GAN training with discriminator D(4) · theoretical analysis | neuron death avoidance | No clear effect |
| Using discriminator weights C_{j,j',r,r'} much smaller for cross-patch terms than within-patch terms (C_{j,j'} << C_{j,j}) | D(5) discriminator for first hidden layer · theoretical analysis | order of learning within-patch vs cross-patch moments | No clear effect |
| Empirical observation: forward super-resolution structure holds in DCGAN trained on LSUN Church; hidden activations become sparse and not-too-positively correlated without explicit regularization | DCGAN generator on LSUN Church · LSUN Church (empirical) | activation sparsity and correlation | Helped |
Closest papers by meaning
- SAN: Inducing Metrizability of GAN with Discriminative Normalized Linear Layer (ICLR 2024 · similarity 0.88)
- GSDD: Generative Space Dataset Distillation for Image Super-resolution (AAAI 2024 · similarity 0.88)
- RG-GAN: Dynamic Regenerative Pruning for Data-Efficient Generative Adversarial Networks (AAAI 2024 · similarity 0.88)
- Uncertainty-Aware GAN for Single Image Super Resolution (AAAI 2024 · similarity 0.87)
- Adversarial Diffusion Compression for Real-World Image Super-Resolution (CVPR 2025 · similarity 0.87)
- SeD: Semantic-Aware Discriminator for Image Super-Resolution (CVPR 2024 · similarity 0.87)
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.