{"id":14958118,"url":"https://github.com/joisino/laf","last_synced_at":"2025-05-07T09:45:01.570Z","repository":{"id":253243345,"uuid":"842915970","full_name":"joisino/laf","owner":"joisino","description":"Code for \"Training-free Graph Neural Networks and the Power of Labels as Features\" (TMLR 2024)","archived":false,"fork":false,"pushed_at":"2024-08-15T11:38:43.000Z","size":192,"stargazers_count":56,"open_issues_count":0,"forks_count":7,"subscribers_count":2,"default_branch":"main","last_synced_at":"2025-03-31T08:44:13.439Z","etag":null,"topics":["deep-learning","gnn","graph-neural-networks","label-propagation","machine-learning","node-classification","pytorch","transductive-learning"],"latest_commit_sha":null,"homepage":"https://arxiv.org/abs/2404.19288","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/joisino.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2024-08-15T11:36:30.000Z","updated_at":"2025-03-30T03:02:09.000Z","dependencies_parsed_at":null,"dependency_job_id":"770d611c-bea3-4500-97a1-0030979e0e13","html_url":"https://github.com/joisino/laf","commit_stats":{"total_commits":2,"total_committers":2,"mean_commits":1.0,"dds":0.5,"last_synced_commit":"8d9a842a247283b2eff166dddb4eb58cb707fac7"},"previous_names":["joisino/laf"],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/joisino%2Flaf","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/joisino%2Flaf/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/joisino%2Flaf/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/joisino%2Flaf/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/joisino","download_url":"https://codeload.github.com/joisino/laf/tar.gz/refs/heads/main","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":252853642,"owners_count":21814554,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["deep-learning","gnn","graph-neural-networks","label-propagation","machine-learning","node-classification","pytorch","transductive-learning"],"created_at":"2024-09-24T13:16:15.274Z","updated_at":"2025-05-07T09:45:01.551Z","avatar_url":"https://github.com/joisino.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Training-free Graph Neural Networks and the Power of Labels as Features (TMLR 2024)\n\n[![arXiv](https://img.shields.io/badge/arXiv-2404.19288-b31b1b.svg)](https://arxiv.org/abs/2404.19288)\n\nWe propose training-free graph neural networks (TFGNNs), which can be used without training and can also be improved with optional training.\n\nPaper: https://arxiv.org/abs/2404.19288\n\n## 💿 Preparation\n\nInstall [Poetry](https://python-poetry.org/) and run the following command:\n\n```bash\n$ poetry install\n```\n\n## 🧪 Evaluation\n\n### TFGNNs Outperform Existing GNNs in Training-free Setting\n\n```bash\n$ poetry run python trainingfree.py\n```\n\n| Model                               | Cora | CiteSeer | PubMed | CS   | Physics | Computers | Photo |\n|-------------------------------------|------|----------|--------|------|---------|-----------|-------|\n| GCNs                                | 0.163 | 0.167    | 0.180  | 0.079 | 0.101   | 0.023     | 0.119 |\n| GCNs + LaF                          | 0.119 | 0.159    | 0.407  | 0.080 | 0.146   | 0.061     | 0.142 |\n| GATs                                | 0.177 | 0.229    | 0.180  | 0.040 | 0.163   | 0.058     | 0.122 |\n| GATs + LaF                          | 0.319 | 0.077    | 0.180  | 0.076 | 0.079   | 0.025     | 0.044 |\n| TFGNNs + random initialization      | 0.149 | 0.177    | 0.180  | 0.023 | 0.166   | 0.158     | 0.090 |\n| **TFGNNs (proposed)**               | **0.600** | **0.362**    | **0.413**  | **0.601** | **0.717**   | **0.730**     | **0.637** |\n\n\n▲ **Results.** TFGNNs outperform existing GNNs in the training-free setting.\n\n### Deep TFGNNs Perform Better in Training-free Setting\n\n```bash\n$ poetry run python numlayers.py\n```\n\n\u003cimg src=\"./imgs/numlayers.png\"\u003e\n\n▲ **Results.** Deep TFGNNs perform better in the training-free setting.\n\n### TFGNNs Converge Fast in Optional Training\n\n```bash\n$ poetry run python training.py\n```\n\n\u003cimg src=\"./imgs/training.png\"\u003e\n\n▲ **Results.** TFGNNs converge fast in optional training.\n\n### TFGNNs are Robust to Feature Noise\n\n```bash\n$ poetry run python noisy.py\n```\n\n\u003cimg src=\"./imgs/noisy.png\"\u003e\n\n▲ **Results.** TFGNNs are robust to feature noise.\n\n\n## 🖋️ Citation\n\n```\n@article{sato2024trainingfree,\n  author    = {Ryoma Sato},\n  title     = {Training-free Graph Neural Networks and the Power of Labels as Features},\n  journal   = {Transactions on Machine Learning Research},\n  year      = {2024},\n}\n```\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fjoisino%2Flaf","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fjoisino%2Flaf","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fjoisino%2Flaf/lists"}