Research map / Label-efficient and robust learning / Semi-supervised learning
Neural Universal Discrete Denoiser
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 |
|---|---|---|---|
| replace DUDE with Neural DUDE (shared network across contexts) | DUDE · Einstein binary image, delta=0.15 | relative BER: 0.578 (5) -> 0.384 (38) | Helped |
| replace DUDE with Neural DUDE | DUDE · Lena binary image, delta=0.1 | relative BER: 0.495 (6) -> 0.403 (38) | Helped |
| replace DUDE with Neural DUDE | DUDE · Barbara binary image, delta=0.15 | relative BER: 0.492 (5) -> 0.448 (33) | Helped |
| replace DUDE with Neural DUDE | DUDE · Cameraman binary image, delta=0.1 | relative BER: 0.310 (5) -> 0.268 (35) | Helped |
| replace DUDE with Neural DUDE | DUDE · Shannon scanned paper, delta=0.1 | relative BER: 0.475 (5) -> 0.402 (35) | Helped |
| increase network depth from 1L (linear softmax) to 4L (ablation) | Neural DUDE · Synthetic BSMC binary data | BER vs optimum: approaches 0.558 optimum; 4L reaches 0.563 | Helped |
| increase context size k | DUDE · Synthetic BSMC binary data | BER | Hurt |
| increase context size k | Neural DUDE (4L) · Synthetic BSMC binary data | BER | No clear effect |
| increase context size k up to 36 on Einstein image | Neural DUDE (4L) · Einstein 256x256, delta=0.1 | BER | Helped |
| choose k by minimizing estimated BER (Algorithm rule) | Neural DUDE (4L) · real binary images and nanopore | BER | Helped |
| use large k (about 100) with Neural DUDE instead of DUDE | DUDE · Simulated nanopore DNA reads | error rate | Helped |
| increase hidden nodes from 40-40-40 to 80-80-80 (ablation) | Neural DUDE (4L) · Simulated nanopore DNA reads | error rate | Helped |
Closest papers by meaning
- Universal Denoising Networks : A Novel CNN Architecture for Image Denoising (CVPR 2018 · similarity 0.89)
- High-Quality Self-Supervised Deep Image Denoising (NeurIPS 2019 · similarity 0.87)
- Self-Supervised Image Denoising Using Implicit Deep Denoiser Prior (AAAI 2023 · similarity 0.87)
- Classic Video Denoising in a Machine Learning World: Robust, Fast, and Controllable (CVPR 2025 · similarity 0.87)
- Noisier2Noise: Learning to Denoise From Unpaired Noisy Data (CVPR 2020 · similarity 0.87)
- Robust And Interpretable Blind Image Denoising Via Bias-Free Convolutional Neural Networks (ICLR 2020 · similarity 0.86)
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.