Exphormer: Sparse Transformers for Graphs

Benchmark Model Rank Results
graph-classification-on-malnet-tinyExphormer#3Accuracy: 94.02±0.209
node-classification-on-amz-photoExphormer#4Accuracy: 95.35±0.22%
node-classification-on-clusterExphormer#5Accuracy: 78.22±0.045
graph-classification-on-mnistExphormer#6Accuracy: 98.414±0.038
node-classification-on-patternExphormer#6Accuracy: 86.74
graph-classification-on-cifar10-100kExphormer#7Accuracy (%): 74.754±0.194
node-classification-on-coco-spExphormer#7macro F1: 0.343±0.0008
node-classification-on-pascalvoc-spExphormer#7macro F1: 0.396±0.0027
node-classification-on-coauthor-physicsExphormer#9Accuracy: 96.89±0.09%
link-prediction-on-pcqm-contactExphormer#10MRR: 0.3637±0.0020
node-classification-on-coauthor-csExphormer#14Accuracy: 94.93±0.46%
graph-regression-on-peptides-structExphormer#19MAE: 0.2481±0.0007
graph-classification-on-peptides-funcExphormer#33AP: 0.6527±0.0043