Table 5 Hyperparameters range for ECS TabNet.
Hyperparameter | Description | Value |
|---|---|---|
\({N_a}\) | Attention layer dimension | {8, 16, 24, 32, 64} |
\({N_d}\) | Decision layer dimension | {8, 16, 24, 32, 64} |
\({N_{step}}\) | Number of decision steps | {3, 4, 5, 6}, |
\(gamma\) | Relaxation factor that controls the mask | {1.0, 1.2, 1.5, 2.0} |
\(momentum\) | Control of the feature selection update process | {0.6, 0.7, 0.8, 0.9} |
