Graph Classification on MUTAG (10-fold cross-validation)
96.4AccuracyTFGW
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| TFGWInput Structure=Adjacency, L=22022.05 | 96.4 | — | |
| ESAN + SSMAaggregation=SSMA2024.09 | 96.32 | — | |
| TFGWInput Structure=Shortest-path, L=22022.05 | 94.8 | — | |
| CANvalidation=10-fold cross-validation2022.09 | 94.1 | — | |
| CIN2021.10 | 92.7 | — | |
| CINvalidation=10-fold cross-validation2022.09 | 92.7 | — | |
| Patchysan2019.05 | 92.6 | — | |
| GINGraph-level pre-training=Supervised, Node-level pre-training=Context Prediction2019.05 | 92.5 | — | |
| GINGraph-level pre-training=None, Node-level pre-training=Context Prediction2019.05 | 92.4 | — | |
| GSN2021.10 | 92.2 | — | |
| GSNvalidation=10-fold cross-validation2022.09 | 92.2 | — | |
| FGSD2019.05 | 92.12 | — | |
| OT-GNNL=42022.05 | 92.1 | — | |
| GINGraph-level pre-training=None, Node-level pre-training=EdgePred2019.05 | 91.9 | — | |
| GINGraph-level pre-training=Supervised, Node-level pre-training=EdgePred2019.05 | 91.9 | — | |
| MGNNIValidation=10-fold cross-validation, Setting=Inductive2022.10 | 91.9 | 5.5 | |
| CCN2019.05 | 91.64 | — | |
| OT-GNNL=22022.05 | 91.6 | — | |
| PATCHY-SAN2022.05 | 91.6 | — | |
| DDGKUnsupervised=true, No Algorithmic Insights=true, No Feature Engineering=true2019.04 | 91.58 | — | |
| MFNBackbone=GCN2023.10 | 91.5 | — | |
| GINGraph-level pre-training=None, Node-level pre-training=Attribute Masking2019.05 | 91.4 | — | |
| HiGCNevaluation_protocol=10-fold cross-validation2023.09 | 91.3 | — | |
| GIN-AK+2021.10 | 91.3 | — | |
| ESANaggregation=sum, reported=true2024.09 | 91.1 | — | |
| GLALabel Ratio=70%2022.05 | 91.05 | — | |
| WEGL2022.05 | 91 | — | |
| GINGraph-level pre-training=Supervised, Node-level pre-training=None2019.05 | 90.9 | — | |
| GINGraph-level pre-training=Supervised, Node-level pre-training=Infomax2019.05 | 90.9 | — | |
| PPGNsevaluation_protocol=10-fold cross-validation2023.09 | 90.6 | — | |
| PPGNComplexity=O(n^3)2021.11 | 90.6 | — | |
| PPGNS2021.10 | 90.6 | — | |
| PPGN2022.05 | 90.6 | — | |
| PPGNsvalidation=10-fold cross-validation2022.09 | 90.6 | — | |
| PPGNN2023.10 | 90.55 | — | |
| Ours 1b=400, d=2, suffix=(ii)2019.05 | 90.55 | — | |
| GIN + SSMAaggregation=SSMA2024.09 | 90.51 | — | |
| WL kernelevaluation_protocol=10-fold cross-validation2023.09 | 90.4 | — | |
| WL subtree2019.05 | 90.4 | — | |
| WL subtreeComplexity=O(n)2021.11 | 90.4 | — | |
| DropGINComplexity=O(rn), r≈202021.11 | 90.4 | — | |
| WL kernel2021.10 | 90.4 | — | |
| WLKvalidation=10-fold cross-validation2022.09 | 90.4 | — | |
| GINGraph-level pre-training=Supervised, Node-level pre-training=Attribute Masking2019.05 | 90.3 | — | |
| RetGK2019.04 | 90.3 | — | |
| RetGKMethod Category=Graph2026.05 | 90.3 | — | |
| GIN2022.05 | 90.1 | — | |
| GLALabel Ratio=50%2022.05 | 90 | — | |
| ECPMethod Category=Topology-based2026.05 | 90 | — | |
| GINMethod Category=Graph2026.05 | 90 | — | |
| PDF2023.10 | 89.91 | — | |
| S2S-N2N-PPUnsupervised=true, No Algorithmic Insights=false, No Feature Engineering=true2019.04 | 89.86 | — | |
| MPSNevaluation_protocol=10-fold cross-validation2023.09 | 89.8 | — | |
| GINGraph-level pre-training=None, Node-level pre-training=Infomax2019.05 | 89.8 | — | |
| DropGIN2022.05 | 89.8 | — | |
| PersLayMethod Category=Topology-based2026.05 | 89.8 | — | |
| GCN + SSMAaggregation=SSMA2024.09 | 89.79 | — | |
| Ours 3b=256, d=3, suffix=(ii)2019.05 | 89.44 | — | |
| GINevaluation_protocol=10-fold cross-validation2023.09 | 89.4 | — | |
| Natural GNevaluation_protocol=10-fold cross-validation2023.09 | 89.4 | — | |
| GIN2023.10 | 89.4 | — | |
| GINSource=Xu et al., 20192019.05 | 89.4 | — | |
| GINComplexity=O(n)2021.11 | 89.4 | — | |
| GIN2021.10 | 89.4 | — | |
| Natural GN2021.10 | 89.4 | — | |
| GINvalidation=10-fold cross-validation2022.09 | 89.4 | — | |
| NGNvalidation=10-fold cross-validation2022.09 | 89.4 | — | |
| CGSValidation=10-fold cross-validation, Setting=Inductive2022.10 | 89.4 | 5.6 | |
| GINevaluation_protocol=10-fold cross-validation2019.05 | 89.4 | — | |
| GIN2019.05 | 89.4 | — | |
| GCN22023.10 | 89.39 | — | |
| GINGraph-level pre-training=None, Node-level pre-training=None2019.05 | 89.3 | — | |
| IGNNValidation=10-fold cross-validation, Setting=Inductive2022.10 | 89.3 | 6.7 | |
| GAT + SSMAaggregation=SSMA2024.09 | 89.19 | — | |
| PatchySanComplexity=O(n)2021.11 | 89 | — | |
| GINValidation=10-fold cross-validation, Setting=Inductive2022.10 | 89 | 6 | |
| PSCN2023.10 | 88.95 | — | |
| PSCNUnsupervised=false, No Algorithmic Insights=false, No Feature Engineering=false2019.04 | 88.95 | — | |
| PSCNk=102019.04 | 88.95 | — | |
| PSCNk=102019.05 | 88.95 | — | |
| EIGNNValidation=10-fold cross-validation, Setting=Inductive2022.10 | 88.9 | 1.1 | |
| Ours 2b=400, d=2, suffix=(i)2019.05 | 88.88 | — | |
| edGNNaggregation=max, runs=102019.04 | 88.8 | — | |
| GATv2 + SSMAaggregation=SSMA2024.09 | 88.8 | — | |
| CapsGNN2023.10 | 88.67 | — | |
| RepSetnode embeddings=struc2vec2019.04 | 88.63 | — | |
| FDGNNValidation=10-fold cross-validation, Setting=Inductive2022.10 | 88.5 | 3.8 | |
| PLACEClassification Protocol=linear SVM, Descriptor Filter=deg+hks102026.05 | 88.4 | — | |
| TabPFN-GNevaluation protocol=10-fold cross-validation2025.12 | 88.36 | — | |
| InfomaxLabel Ratio=70%2022.05 | 88.33 | — | |
| SimGRACELabel Ratio=70%2022.05 | 88.33 | — | |
| JOAOv2Label Ratio=70%2022.05 | 88.33 | — | |
| WKPI-kCMethod Category=Topology-based2026.05 | 88.3 | — | |
| PNA + SSMAaggregation=SSMA2024.09 | 88.29 | — | |
| MVGRLLabel Ratio=70%2022.05 | 87.95 | — | |
| MLGClassifier=binary C-SVM, Kernel levels=l in {1, 2, 3, 4}, Radius size=n in {1, 2, 3, 4}2016.03 | 87.94 | 1.61 | |
| AWE2023.10 | 87.87 | — | |
| AWE-FB2019.05 | 87.87 | — | |
| Set-Transformernode embeddings=struc2vec2019.04 | 87.71 | — | |
| NN-meannode embeddings=struc2vec2019.04 | 87.55 | — |