Research map / Graph learning / GNN architectures
SigMaNet: One Laplacian to Rule Them All
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 SigMaNet's Sign-Magnetic Laplacian instead of the strongest reported alternatives | link sign prediction on Bitcoin Alpha, Bitcoin-OTC, and WikiRfa · Bitcoin Alpha, Bitcoin-OTC, WikiRfa | AUC: 89.88 -> 92.46; 91.52 -> 93.67; 89.81 -> 90.53 | Helped |
| Use SigMaNet instead of the strongest reported alternatives | link sign prediction on Slashdot · Slashdot | AUC: 90.17 -> 89.79 | Hurt |
| Use SigMaNet instead of the strongest reported alternatives | existence prediction · Bitcoin Alpha*, Bitcoin-OTC*, Telegram, and DSBM graphs | accuracy: 83.26 -> 83.28; 84.77 -> 84.71; 86.32 -> 84.95; 61.27 -> 62.25; 63.81 -> 64.48; 64.93 -> 65.49 | Mixed |
| Use SigMaNet instead of the strongest reported alternatives | direction prediction · Bitcoin-OTC* and DSBM graphs | accuracy: 56.42 -> 57.19; 68.50 -> 69.10; 72.01 -> 72.74; 73.28 -> 73.77 | Helped |
| Use SigMaNet instead of the strongest reported alternatives | direction prediction · Telegram and Bitcoin Alpha* | accuracy: 91.65 -> 91.20; 58.62 -> 56.90 | Hurt |
| Use SigMaNet instead of the strongest reported alternatives | node classification · Telegram | accuracy: 72.69 -> 74.23 | Helped |
| Use SigMaNet instead of the strongest reported alternatives | node classification · DSBM αij=0.05, 0.08, and 0.1 | accuracy: 91.04 -> 87.44; 94.94 -> 96.14; 96.77 -> 98.60 | Mixed |
| Use SigMaNet instead of the strongest reported alternatives | link sign prediction on unit-weight graphs · Slashdot and Epinions | Micro-F1: 86.16 -> 85.03; 93.55 -> 92.25 | Hurt |
| Retain edge direction and weight information rather than symmetrizing it away | SigMaNet node classification · synthetic DSBM graphs | accuracy | Helped |
| Use a positively homogeneous Sign-Magnetic Laplacian rather than the Magnetic Laplacian | SigMaNet on graphs with large edge weights · synthetic DSBM node classification | accuracy: MagNet 78.64$\pm$1.29 -> SigMaNet 87.44$\pm$0.99; MagNet 87.52$\pm$1.30 -> SigMaNet 96.14$\bm{\pm}$0.64; MagNet 91.58$\p | Helped |
Closest papers by meaning
- Graph Learning in 4D: A Quaternion-Valued Laplacian to Enhance Spectral GCNs (AAAI 2024 · similarity 0.91)
- Signed Laplacian Graph Neural Networks (AAAI 2023 · similarity 0.90)
- Laplacian Canonization: A Minimalist Approach to Sign and Basis Invariant Spectral Embedding (NeurIPS 2023 · similarity 0.89)
- WaveNet: Tackling Non-stationary Graph Signals via Graph Spectral Wavelets (AAAI 2024 · similarity 0.88)
- HoloNets: Spectral Convolutions do extend to Directed Graphs (ICLR 2024 · similarity 0.88)
- Position: Spectral GNNs Rely Less on Graph Fourier Basis than Conceived (ICML 2025 · similarity 0.88)
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.