Research map / Architectures and efficiency / RNNs and normalization
Boosting Learning for LDPC Codes to Improve the Error-Floor Performance
Hee-Youl Kwak, Dae-Young Yun, Yongjune Kim, Sang-Hyo Kim, Jong-Seon No
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 |
|---|---|---|---|
| Train the post decoder on error-floor uncorrected words rather than unfiltered received words | NMS decoder · WiMAX LDPC, error-floor samples at 4.5 dB | Test FER: 0.315 times of the FER after base decoding | Improved |
| Train the post decoder on uncorrected words from the waterfall region instead of the error-floor region | NMS decoder · WiMAX LDPC, waterfall samples at 3.5 dB | Test FER: 0.77 times of the FER after base decoding | No clear effect |
| Train the post decoder on unfiltered received words | NMS decoder · WiMAX LDPC, error-floor region | FER | Hurt |
| Add boosting with uncorrected words (ablation) | NMS decoder · WiMAX LDPC, Eb/N0 5.0dB | FER: 3.14×10^{-6} -> 2.77×10^{-7} | Improved |
| Add block-wise training to boosting (ablation) | NMS decoder · WiMAX LDPC, Eb/N0 5.0dB | FER: 2.77×10^{-7} -> 1.84×10^{-7} | Improved |
| Use proposed UCN/SCN weight sharing instead of full diversity weights (ablation) | NMS decoder · WiMAX LDPC, Eb/N0 5.0dB | FER: 1.84×10^{-7} -> 1.85×10^{-7} | No clear effect |
| Use block-wise training with retraining rather than one-shot training (ablation) | NMS decoder · WiMAX LDPC, post decoder, 50 iterations | Test FER: 0.18 -> 0.11 | Improved |
| Use block-wise training with retraining rather than iter-by-iter training (ablation) | NMS decoder · WiMAX LDPC, post decoder, 50 iterations | Test FER: 0.37 -> 0.11 | Improved |
| Add retraining with Δ1=5 and Δ2=10 rather than no retraining with Δ1=5 (ablation) | NMS decoder · WiMAX LDPC, block-wise parameter search | Test FER: 0.193 -> 0.112 | Improved |
| Increase block size from Δ1=1 to Δ1=10 with no retraining (ablation) | NMS decoder · WiMAX LDPC, block-wise parameter search | Test FER: 0.376 -> 0.142 | Improved |
| Replace boosting with a single large network of the same model size (ablation) | NMS decoder · WiMAX LDPC, error-floor region | FER | Hurt |
| Use spatial or temporal sharing instead of dynamic UCN/SCN weights (ablation) | NMS decoder · WiMAX LDPC, error-floor region | FER | Hurt |
Closest papers by meaning
- Learning Linear Block Error Correction Codes (ICML 2024 · similarity 0.84)
- Friendly Attacks to Improve Channel Coding Reliability (AAAI 2024 · similarity 0.83)
- LK Losses: Direct Acceptance Rate Optimization for Speculative Decoding (ICML 2026 · similarity 0.80)
- AdaDecode: Accelerating LLM Decoding with Adaptive Layer Parallelism (ICML 2025 · similarity 0.79)
- DMax: Aggressive Parallel Decoding for dLLMs (NeurIPS 2026 · similarity 0.79)
- DeepPolar: Inventing Nonlinear Large-Kernel Polar Codes via Deep Learning (ICML 2024 · similarity 0.79)
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.
