Research map / Image generation / Autoregressive and latent generation
Images as Weight Matrices: Sequential Image Generation Through Synaptic Learning Rules
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 |
|---|---|---|---|
| Add a U-Net refinement stage | FPA · CelebA | FID: 18.3 -> 3.7 | Helped |
| Add a U-Net refinement stage | FPA · MetFaces | FID: 36.3 -> 24.5 | Helped |
| Add a U-Net refinement stage | FPA · AFHQ-Cat | FID: 17.1 -> 6.8 | Helped |
| Add a U-Net refinement stage | FPA · AFHQ-Dog | FID: 45.3 -> 19.9 | Helped |
| Add a U-Net refinement stage | FPA · AFHQ-Wild | FID: 20.2 -> 9.4 | Helped |
| Add a U-Net refinement stage | FPA · LSUN-Church | FID: 42.8 -> 5.2 | Helped |
| Reduce painting steps from 64 to 32 (ablation) | FPA · MetFaces | FID: 36.3 -> 43.6 | Hurt |
| Reduce painting steps from 64 to 16 (ablation) | FPA · MetFaces | FID: 36.3 -> 53.8 | Hurt |
| Reduce painting steps from 64 to 8 (ablation) | FPA · MetFaces | FID: 36.3 -> 80.1 | Hurt |
| Reduce painting steps from 64 to 32 (ablation) | FPA · AFHQ-Cat | FID: 17.1 -> 25.9 | Hurt |
| Reduce painting steps from 64 to 16 or 8 (ablation) | FPA · AFHQ-Cat | FID: 17.1 -> 86.6; 17.1 -> 164.4 | Hurt |
| Reduce painting steps from 64 to 32, 16, or 8 (ablation) | FPA · AFHQ-Dog | FID: 45.3 -> 69.5; 45.3 -> 165.8; 45.3 -> 220.1 | Hurt |
Closest papers by meaning
- A Powerful Generative Model Using Random Weights for the Deep Image Representation (NeurIPS 2016 · similarity 0.88)
- Generative Modeling of Weights: Generalization or Memorization? (CVPR 2026 · similarity 0.88)
- Plug & Play Generative Networks: Conditional Iterative Generation of Images in Latent Space (CVPR 2017 · similarity 0.88)
- Enhancing Consistency-Based Image Generation via Adversarialy-Trained Classification and Energy-Based Discrimination (NeurIPS 2024 · similarity 0.85)
- HyperNet Fields: Efficiently Training Hypernetworks without Ground Truth by Learning Weight Trajectories (CVPR 2025 · similarity 0.85)
- NitroFusion: High-Fidelity Single-Step Diffusion through Dynamic Adversarial Training (CVPR 2025 · 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.