Table 7:
CTDT Hyperparameters
| Synthetic Data | Kidney | HIV | |
|---|---|---|---|
| Training Steps | 1E5 | 1E5 | 7500 |
| Batch Size | 64 | 64 | 4 |
| Value Embedding y Dimension | 128 | 128 | 96 |
| Type embedding e Dimension | 64 | 64 | 32 |
| Temporal Embedding t Dimension | 64 | 64 | 32 |
| Number Of Layers | 3 | 3 | 4 |
| Number Of Attention Heads | 1 | 1 | 4 |
| Learning Rate | 0.0001 | 0.0001 | 0.006 |
| Attention Window Size | 20 | 20 | unlimited |