Table 2 Hyper-parameter settings for each dataset.
From: Graph representation learning via enhanced GNNs and transformers
Hyperparameter | ZINC-12K | CIFAR10 | PATTERN | CLUSTER | MNIST | ogbg-molhiv | ogbg-molpcba | Peptides-func | Peptides-struct |
|---|---|---|---|---|---|---|---|---|---|
Transformer Layers | 10 | 3 | 10 | 16 | 3 | 10 | 5 | 4 | 4 |
Hidden dim | 64 | 52 | 64 | 48 | 52 | 64 | 384 | 96 | 96 |
Heads | 8 | 4 | 8 | 4 | 4 | 4 | 4 | 4 | 8 |
Dropout | 0 | 0 | 0 | 0.01 | 0 | 0.05 | 0.3 | 0 | 0.05 |
Attention dropout | 0.2 | 0.5 | 0.2 | 0.5 | 0.5 | 0.5 | 0.5 | 0.5 | 0.5 |
Graph pooling | Sum | Mean | – | Mean | Mean | Mean | Mean | Mean | Mean |
Epochs | 2000 | 200 | 100 | 100 | 200 | 100 | 100 | 200 | 200 |
