Awesome-Graph-Scaling
April 10, 2025 · View on GitHub
Scope of Graph Scaling
Given that graph data consists of a massive number of nodes and their relationships, Graph Scaling (GS) solves the problem of: How to condense large-scale graphs into smaller yet informative ones.
How can this repository be of service
This repository contains a list of papers who shares a common motivation of GS, including methods of graph condensation, reduction, summarization, etc.; We categorize them based on their aspect of making scaled graphs informative, i.e., what information of the original graph was designed to preserve, the graph properties (graph guided) or the trained models' capabilities (model guided). The detailed statistic of the commonly used datasets can be found after the paper list. Previously, this repository contained a paper list of Graph Condensation, which has been merged within the scope of graph scaling. The previous version can be found at the bottom of this file.
We will make this list updated. If you found any error or any missed paper, please don't hesitate to open an issue or pull request.
Paper List
:triangular_flag_on_post: Our survey paper Learning to Reduce the Scale of Large Graphs: A Comprehensive Survey has been accepted for publication in the Transactions on Knowledge Discovery from Data. We are looking forward to any comments or discussions on this topic :)
Benchmark Dataset Statics
In order to facilitate further experimental research in the field, we have compiled a dataset of commonly used GS methods, with links to jump directly to the specific methods.
Single Graph
| Dataset | #Nodes | #Edges | #Features | #Classes | Type |
|---|---|---|---|---|---|
| Cora | 2,708 | 5,429 | 1,433 | 7 | Citation |
| CoraFull | 19,793 | 130,622 | 8,710 | 70 | Citation |
| Citeseer | 3,327 | 4,732 | 3,703 | 6 | Citation |
| Citeseer-L | 3,327 | 4,732 | 3,703 | 2 | Citation |
| DBLP | 17,716 | 52,867 | 1,639 | 4 | Citation |
| DBLP-large | 26,128 | 105,734 | 4,057 | 2 | Citation |
| Coauthor Physics(Co-phy) | 34,493 | 247,692 | 8,415 | 5 | Citation |
| OGBNArxiv(Arixiv) | 169,343 | 1,166,243 | 128 | 40 | Citation |
| Pubmed | 19,717 | 44,338 | 500 | 3 | Citation |
| OGBLCitation2 | 2,927,963 | 30,561,187 | 128 | N/A | Citation |
| ACM | 3,025 | 13,128 | 1,870 | N/A | Citation |
| Friendster | 7,944,949 | 446,673,688 | N/A | 5000 | Social |
| Flickr | 89,250 | 899,756 | 500 | 7 | Social |
| 232,965 | 57,307,946 | 602 | 210 | Social | |
| Reddit-A | 232,965 | 11,606,919 | 602 | 41 | Social |
| Reddit-B | 227,853 | 114,615,892 | 602 | 40 | Social |
| Genius | 42,1961 | 984,979 | 12 | 2 | Social |
| YelpChi | 45,954 | 3,846,979 | 32 | 2 | Social |
| Polblogs | 1,490 | 16,715 | 5,000 | N/A | Social |
| PPI | 14,755 | 222,055 | 50 | 121 | Biology |
| Yeast | 2,361 | 13,292 | 5,000 | N/A | Biology |
| Airfoil | 4,253 | 12,289 | 5,000 | N/A | Transportation |
| Minnesota | 2,642 | 3,304 | 5000 | N/A | Transportation |
| Bunny | 2,503 | 78,292 | 5,000 | N/A | Point cloud |
| OGBLCollab | 235,868 | 1,285,465 | 128 | N/A | Collaboration |
| OGBNProducts | 2,449,029 | 61,859,140 | 128 | 47 | Product |
| Products | 2,449,029 | 61,859,140 | 100 | 46 | Product |
| Amazon | 11,944 | 4,398,392 | 25 | 2 | Product |
Multiple Graph
| Dataset | #Graphs | #Avg.Nodes | #Avg.Edges | #Classes | Type |
|---|---|---|---|---|---|
| CIFAR10 | 60,000 | 117.6 | 941.07 | 10 | Superpixel |
| ogbg-molhiv | 41,127 | 25.5 | 54.9 | 2 | Molecule |
| ogbg-molbace | 1,513 | 34.1 | 36.9 | 2 | Molecule |
| ogbg-molbbbp | 2,039 | 24.1 | 26.0 | 2 | Molecule |
| MUTAG | 188 | 17.93 | 19.79 | 2 | Molecule |
| NCI1 | 4,110 | 29.87 | 32.30 | 2 | Molecule |
| DD | 1,178 | 284.32 | 715.66 | 2 | Molecule |
| ENZYMES | 600 | 32.63 | 62.14 | 6 | Molecule |
| NCI109 | 4,127 | 29.68 | 32.13 | 2 | Molecule |
| PROTEINS | 1,108 | 39.06 | 72.70 | 2 | Molecule |
| PTC | 344 | 14.29 | 14.69 | 2 | Molecule |
| IMDB-BINARY | 1,000 | 19.8 | 96.5 | 2 | Social |
| IMDB-MULTI | 1,500 | 13.0 | 65.9 | 3 | Social |
| E-commerce | 1,109 | 33.7 | 46.3 | 2 | Transaction |
Previous Version of this repo
Awesome-Graph-Condensation
:triangular_flag_on_post: We have released a new survey paper, presenting a comprehensive overview of existing graph condensation methods. We are looking forward to any comments or discussions on this topic :)
What is GC
Given that graph data consists of a massive number of nodes and their relationships, Graph Condensation (GC) solves the problem of: How to condense large-scale graphs into smaller yet informative ones.
How can this repository be of service
This repository contains a list of papers who shares a common motivation of GC; We categorize them based on their aspect of making condensed graphs informative, i.e., what information of the original graph was designed to preserve, the graph properties (graph guided) or the trained models' capabilities (model guided).
We will try to make this list updated. If you found any error or any missed paper, please don't hesitate to open an issue or pull request.
Paper List
| Survey Paper | Conference |
|---|---|
| :triangular_flag_on_post: A Survey on Graph Condensation | arXiv 2024 |
| Graph Condensation: A Survey | arXiv 2024 |
| A Comprehensive Survey on Graph Reduction: Sparsification, Coarsening, and Condensation | arXiv 2024 |