| Model . | No. of parameters . | Path (ms) . | Training (ms) . | Inference (ms) . | RMSE . |
|---|---|---|---|---|---|
| GCN | 519 676 | 21.73 ± 10.44 | 21.04 ± 10.84 | 0.37 ± 0.02 | |
| GAT | 523 768 | 23.84 ± 13.25 | 24.99 ± 11.41 | 0.36 ± 0.02 | |
| GraphSAGE | 1 035 260 | 18.77 ± 11.06 | 16.97 ± 12.16 | 0.38 ± 0.01 | |
| Transformer | 2 078 704 | 33.94 ± 16.16 | 38.13 ± 8.71 | 0.36 ± 0.01 | |
| Impeller | 538 108 | 1.61 ± 0.60 | 6.35 ± 0.30 | 8.43 ± 0.25 | 0.34 ± 0.00 |
| Model . | No. of parameters . | Path (ms) . | Training (ms) . | Inference (ms) . | RMSE . |
|---|---|---|---|---|---|
| GCN | 519 676 | 21.73 ± 10.44 | 21.04 ± 10.84 | 0.37 ± 0.02 | |
| GAT | 523 768 | 23.84 ± 13.25 | 24.99 ± 11.41 | 0.36 ± 0.02 | |
| GraphSAGE | 1 035 260 | 18.77 ± 11.06 | 16.97 ± 12.16 | 0.38 ± 0.01 | |
| Transformer | 2 078 704 | 33.94 ± 16.16 | 38.13 ± 8.71 | 0.36 ± 0.01 | |
| Impeller | 538 108 | 1.61 ± 0.60 | 6.35 ± 0.30 | 8.43 ± 0.25 | 0.34 ± 0.00 |
Training and inference times with the fastest performance, as well as the best imputation performance (RMSE), are highlighted in bold.
| Model . | No. of parameters . | Path (ms) . | Training (ms) . | Inference (ms) . | RMSE . |
|---|---|---|---|---|---|
| GCN | 519 676 | 21.73 ± 10.44 | 21.04 ± 10.84 | 0.37 ± 0.02 | |
| GAT | 523 768 | 23.84 ± 13.25 | 24.99 ± 11.41 | 0.36 ± 0.02 | |
| GraphSAGE | 1 035 260 | 18.77 ± 11.06 | 16.97 ± 12.16 | 0.38 ± 0.01 | |
| Transformer | 2 078 704 | 33.94 ± 16.16 | 38.13 ± 8.71 | 0.36 ± 0.01 | |
| Impeller | 538 108 | 1.61 ± 0.60 | 6.35 ± 0.30 | 8.43 ± 0.25 | 0.34 ± 0.00 |
| Model . | No. of parameters . | Path (ms) . | Training (ms) . | Inference (ms) . | RMSE . |
|---|---|---|---|---|---|
| GCN | 519 676 | 21.73 ± 10.44 | 21.04 ± 10.84 | 0.37 ± 0.02 | |
| GAT | 523 768 | 23.84 ± 13.25 | 24.99 ± 11.41 | 0.36 ± 0.02 | |
| GraphSAGE | 1 035 260 | 18.77 ± 11.06 | 16.97 ± 12.16 | 0.38 ± 0.01 | |
| Transformer | 2 078 704 | 33.94 ± 16.16 | 38.13 ± 8.71 | 0.36 ± 0.01 | |
| Impeller | 538 108 | 1.61 ± 0.60 | 6.35 ± 0.30 | 8.43 ± 0.25 | 0.34 ± 0.00 |
Training and inference times with the fastest performance, as well as the best imputation performance (RMSE), are highlighted in bold.
This PDF is available to Subscribers Only
View Article Abstract & Purchase OptionsFor full access to this pdf, sign in to an existing account, or purchase an annual subscription.
