Research map / Image generation / Image restoration
From Denoising to De-Channeling: Integrating Physical Channel Priors into Diffusion Models for Radio Signal Understanding
Yaoqi Liu, Jin Wang, Chunchen Wang, Hui Wang, Chuan Shi
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 |
|---|---|---|---|
| add diffusion pretraining (vs fine-tuning FusedFormer directly) (ablation) | PWC-Diff · RML2022 (avg over SNR) | top-1 accuracy: 51.22 -> 69.97 | Improved |
| add diffusion pretraining (ablation) | PWC-Diff · RML2016.10A (avg over SNR) | top-1 accuracy: 57.77 -> 63.60 | Improved |
| add diffusion pretraining (ablation) | PWC-Diff · TechRec (avg over SNR) | top-1 accuracy: 90.25 -> 91.02 | Improved |
| add physical channel priors in pretraining instead of Gaussian-only noise (ablation) | PWC-Diff · RML2022 (avg over SNR) | top-1 accuracy: 64.30 -> 69.97 | Improved |
| add physical channel priors in pretraining instead of Gaussian-only noise (ablation) | PWC-Diff · RML2016.10A (avg over SNR) | top-1 accuracy: 62.86 -> 63.60 | Improved |
| add physical channel priors in pretraining instead of Gaussian-only noise (ablation) | PWC-Diff · TechRec (avg over SNR) | top-1 accuracy: 89.21 -> 91.02 | Improved |
| de-channel (remove channel effects) rather than add channel effects during fine-tuning (ablation) | PWC-Diff · TechRec (avg over SNR) | top-1 accuracy: 85.60 -> 91.02 | Improved |
| de-channel rather than add channel effects during fine-tuning (ablation) | PWC-Diff · RML2022 (avg over SNR) | top-1 accuracy: 66.30 -> 69.97 | Improved |
| include cfo in candidate channel set (ablation) | PWC-Diff · RML2016.10A (avg over SNR) | top-1 accuracy: 61.46 -> 63.60 | Improved |
| include sro in candidate channel set (ablation) | PWC-Diff · RML2016.10A (avg over SNR) | top-1 accuracy: 61.38 -> 63.60 | Improved |
| include rician in candidate channel set (ablation) | PWC-Diff · RML2022 (avg over SNR) | top-1 accuracy: 68.38 -> 69.97 | Improved |
| include multipath in candidate channel set (ablation) | PWC-Diff · RML2022 (avg over SNR) | top-1 accuracy: 68.69 -> 69.97 | Improved |
Closest papers by meaning
- Physics-Informed Generative Modeling of Wireless Channels (ICML 2025 · similarity 0.84)
- NeWRF: A Deep Learning Framework for Wireless Radiation Field Reconstruction and Channel Prediction (ICML 2024 · similarity 0.83)
- Denoising Task Routing for Diffusion Models (ICLR 2024 · similarity 0.82)
- Physics Unrolled Neural Operator for Wireless Field Modeling (NeurIPS 2026 · similarity 0.81)
- Why DiT Models Underperform as Representation Learners without Long Skip Connections (NeurIPS 2026 · similarity 0.81)
- Deconstructing Denoising Diffusion Models for Self-Supervised Learning (ICLR 2025 · similarity 0.81)
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.
