Research map / Label-efficient and robust learning / Self-supervised and contrastive
Feedback Control for Multi-Objective Graph Self-Supervision
Karish Grover, Theodore Vasiloudis, Han Xie, Sixing Lu, Xiang song, Christos Faloutsos
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 |
|---|---|---|---|
| add the log-hypervolume planner (vs uniform allocation) (ablation) | ControlG · Cora node classification | accuracy: 78.52 -> 81.92 | Improved |
| add the log-hypervolume planner (vs uniform allocation) (ablation) | ControlG · PubMed node classification | accuracy: 80.86 -> 84.24 | Improved |
| add both state signals (spectral demand and interference) (ablation) | ControlG · Arxiv node classification | accuracy: 69.82 -> 72.86 | Improved |
| add spectral demand signal (ablation) | ControlG · 9 node classification datasets | accuracy: Cora 80.14 -> 81.92 | Improved |
| add interference signal (ablation) | ControlG · Chameleon, Squirrel | accuracy: Chameleon 68.26 -> 69.54; Squirrel 52.12 -> 53.18 | Improved |
| add PID deficit-tracking controller (vs i.i.d. sampling from f) (ablation) | ControlG · node classification | accuracy: Cora 80.28 -> 81.92 | Improved |
| use single-task temporal blocks with random scheduling instead of per-step gradient mixing | Random scheduler vs ParetoGNN/PCGrad/CAGrad/AutoSSL · node clustering | average rank: avg rank 5.0 vs 7.3-8.4 | Improved |
| add feedback-controlled scheduling on top of random temporal scheduling | Random scheduler · node classification | average rank: avg rank 9.4 -> 1.4 | Improved |
| use ControlG instead of per-step mixing methods | AUTOSSL, ParetoGNN · training time | runtime: 5-15x faster | Improved |
| replace per-step Pareto mixing with ControlG | ParetoGNN · PubMed node classification | accuracy: 67.72 -> 84.24 | Improved |
Closest papers by meaning
- From Coarse to Fine: Enable Comprehensive Graph Self-supervised Learning with Multi-granular Semantic Ensemble (ICML 2024 · similarity 0.85)
- Exploring Correlations of Self-Supervised Tasks for Graphs (ICML 2024 · similarity 0.84)
- Revisiting Self-Supervised Heterogeneous Graph Learning from Spectral Clustering Perspective (NeurIPS 2024 · similarity 0.84)
- Every Node Is Different: Dynamically Fusing Self-Supervised Tasks for Attributed Graph Clustering (AAAI 2024 · similarity 0.84)
- BANGS: Game-theoretic Node Selection for Graph Self-Training (ICLR 2025 · similarity 0.83)
- Uncovering the Redundancy in Graph Self-supervised Learning Models (NeurIPS 2024 · similarity 0.83)
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.
