Research map / Graph learning / Graph algorithms and clustering
Generalized Matrix Means for Semi-Supervised Learning with Multilayer Graphs
Pedro Mercado, Francesco Tudisco, Matthias Hein
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 |
|---|---|---|---|
| Use the power mean Laplacian with a smaller power parameter (L-10 rather than L10) (ablation) | Power Mean Laplacian SSL · Two-layer SBM with one informative layer and a potentially noisy second layer | classification error | Improved |
| Use smaller power parameters for the Laplacian mean (ablation) | Power Mean Laplacian SSL · Complementary-information SBM | test error | Improved |
| Combine information from all three layers using the power mean Laplacian | Power Mean Laplacian SSL · Three-layer, three-class information-independent SBM | classification error | Improved |
| Use L-10 instead of the arithmetic-mean Laplacian L1 (ablation) | Power Mean Laplacian SSL · Three-layer, three-class information-independent SBM | classification error | Improved |
| Apply the modified objective with class-dependent scaling C to address unequal numbers of labeled nodes (ablation) | Power Mean Laplacian SSL · Unbalanced class proportions in labeled data under SBM | average test classification error | Improved |
| Use Class Mass Normalization instead of the proposed class-scaled objective (ablation) | Power Mean Laplacian SSL · Unbalanced class proportions in labeled data under SBM | average test classification error | Hurt |
| Use L-1 rather than the strongest listed baseline at 5% labels | Multilayer graph SSL · BBC | test error: 19.3 -> 17.0 | Improved |
| Use L-1 rather than the strongest listed baseline at 5% labels | Multilayer graph SSL · BBC Sports | test error: 17.4 -> 11.6 | Improved |
| Use L-1 rather than the strongest listed baseline at 5% labels | Multilayer graph SSL · Wikipedia | test error: 56.8 -> 52.3 | Improved |
| Use L-1 rather than the strongest listed baseline at 5% labels | Multilayer graph SSL · CiteSeer | test error: 32.3 -> 39.0 | Hurt |
| Use L-1 rather than the strongest listed baseline at 5% labels | Multilayer graph SSL · Cora | test error: 27.7 -> 31.8 | Hurt |
Closest papers by meaning
- Large-Scale Graph-Based Semi-Supervised Learning via Tree Laplacian Solver (AAAI 2016 · similarity 0.88)
- Learning the Right Layers a Data-Driven Layer-Aggregation Strategy for Semi-Supervised Learning on Multilayer Graphs (ICML 2023 · similarity 0.87)
- A Flexible Generative Framework for Graph-based Semi-supervised Learning (NeurIPS 2019 · similarity 0.87)
- A Random Matrix Analysis and Improvement of Semi-Supervised Learning for Large Dimensional Data (JMLR 2018 · similarity 0.87)
- Continuous Partitioning for Graph-Based Semi-Supervised Learning (NeurIPS 2024 · similarity 0.85)
- Normalize Then Propagate: Efficient Homophilous Regularization for Few-Shot Semi-Supervised Node Classification (AAAI 2025 · similarity 0.85)
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.
