Research map / Graph learning / GNN architectures
Hyper-Graph-Network Decoders for Block Codes
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 |
|---|---|---|---|
| Use hypernetwork-generated adaptive variable-node weights (ablation) | Hyper-Graph-Network decoder · BCH (31,16), 5 iterations, 4 dB | negative natural logarithm of BER: 2.94 -> 4.96 | Helped |
| Increase g's capacity to match the combined parameter count but omit the hypernetwork (ablation) | Hyper-Graph-Network decoder · BCH (31,16), 5 iterations, 4 dB | negative natural logarithm of BER: 4.96 -> 2.94 | Hurt |
| Feed absolute message values to f instead of signed messages (ablation) | Hyper-Graph-Network decoder · BCH (31,16), 5 iterations, 4 dB | negative natural logarithm of BER: 2.86 -> 4.96 | Helped |
| Use the Taylor approximation instead of vanilla arctanh (ablation) | Hyper-Graph-Network decoder · BCH (31,16), 5 iterations, 4 dB | negative natural logarithm of BER: 0.69 -> 4.96 | Helped |
| Add gradient clipping to the vanilla arctanh variant (ablation) | Hyper-Graph-Network decoder · BCH (31,16), 5 iterations, 4 dB | negative natural logarithm of BER: 0.69 -> 0.69 | No clear effect |
| Use a deeper hypernetwork f with 128 neurons per layer (ablation) | Hyper-Graph-Network decoder · BCH (63,51), 5 iterations, 4 dB | negative natural logarithm of BER: 4.64 -> 4.67 | Helped |
Closest papers by meaning
- Cyclically Equivariant Neural Decoders for Cyclic Codes (ICML 2021 · similarity 0.86)
- Learning Linear Block Error Correction Codes (ICML 2024 · similarity 0.85)
- Friendly Attacks to Improve Channel Coding Reliability (AAAI 2024 · similarity 0.82)
- Higher-Order Graphon Neural Networks: Approximation and Cut Distance (ICLR 2025 · similarity 0.82)
- Verifying message-passing neural networks via topology-based bounds tightening (ICML 2024 · similarity 0.81)
- Recurrent Graph Neural Networks and Their Connections to Bisimulation and Logic (AAAI 2024 · similarity 0.80)
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.