Research map / Image generation / Autoregressive and latent generation
Conditional Generation Using Polynomial Expansions
Grigorios Chrysos, Markos Georgopoulos, Yannis Panagakis
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 |
|---|---|---|---|
| replace SNGAN class conditioning with CoPE two-variable polynomial | SNGAN generator · CIFAR10 class-conditional | FID: 14.70 ˘ 0.97 -> 14.22 ˘ 0.66 | Improved |
| replace SNGAN class conditioning with CoPE | SNGAN generator · CIFAR10 class-conditional | IS: 8.30 ˘ 0.11 -> 8.77 ˘ 0.12 | Improved |
| concatenate the class input into SNGAN (SNGAN-CONC) | SNGAN generator · CIFAR10 class-conditional | FID: 14.70 ˘ 0.97 -> 30.65 ˘ 3.55 | Hurt |
| use SPADE-style conditioning in SNGAN | SNGAN generator · CIFAR10 class-conditional | FID: 14.70 ˘ 0.97 -> 21.74 ˘ 0.73 | Hurt |
| replace Pi-Net with CoPE | Pi-Net generator · CIFAR10 class-conditional | IS: 7.54 ˘ 0.16 -> 7.87 ˘ 0.21 | Improved |
| replace Pi-Net with CoPE | Pi-Net generator · CIFAR10 class-conditional | FID: 37.26 ˘ 1.86 -> 34.35 ˘ 2.68 | Improved |
| replace Pi-Net with CoPE | Pi-Net generator · Cars196 class-conditional | FID: 120.40 ˘ 28.65 -> 55.48 ˘ 3.16 | Improved |
| use products of polynomials (Pi-Net) instead of single-variable concatenation (Pi-Net-SICONC) | Pi-Net generator · CIFAR10 class-conditional | FID: 71.81 ˘ 33.00 -> 37.26 ˘ 1.86 | Improved |
| replace SPADE with CoPE (adds products of polynomials) | Pi-Net generator · Cars196 class-conditional | FID: 168.19 ˘ 39.71 -> 55.48 ˘ 3.16 | Improved |
| replace Pi-Net-SICONC with CoPE | Pi-Net generator · Cars196 super-resolution 8x | FID: 186.42 ˘ 40.84 -> 62.76 ˘ 4.37 | Improved |
| replace SPADE with SPADE-CoPE | SPADE · Cars196 super-resolution 16x | FID: 111.75 ˘ 13.41 -> 72.63 ˘ 3.18 | Improved |
| replace SPADE-CoPE with full two-variable CoPE | SPADE-CoPE · Cars196 super-resolution 16x | FID: 72.63 ˘ 3.18 -> 60.42 ˘ 6.19 | Improved |
Closest papers by meaning
- A Simple Approach to Unifying Diffusion-based Conditional Generation (ICLR 2025 · similarity 0.85)
- Guiding a Diffusion Model with a Bad Version of Itself (NeurIPS 2024 · similarity 0.84)
- Return of Unconditional Generation: A Self-supervised Representation Generation Method (NeurIPS 2024 · similarity 0.84)
- Referee Can Play: An Alternative Approach to Conditional Generation via Model Inversion (ICML 2024 · similarity 0.84)
- REG: Rectified Gradient Guidance for Conditional Diffusion Models (ICML 2025 · similarity 0.84)
- Manifold Preserving Guided Diffusion (ICLR 2024 · similarity 0.84)
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.
