TABLE VIII.
Optimal Hyperparameters for LINCS Full Primary Site
| GCNN | regularization | 3.08e–3 |
| num_epochs | 350 | |
| Fs | [[41]] | |
| batch_size | 68 | |
| M | [135, 12] | |
| ps | [[2]] | |
| decay_steps | 380 | |
| momentum | 9.45e–1 | |
| learning_rate | 3.13e–3 | |
| pool | apool1 | |
| decay_rate | 9.89e–1 | |
| Ks | [[5]] | |
| dropout | 5.62e–1 | |
| FF-ANN | learning_rate_init | 5.53e–2 |
| activation | relu | |
| momentum | 8.67e–1 | |
| nesterovs_momentum | True | |
| learning_rate | invscaling | |
| power_t | 2.26e–1 | |
| early_stopping | False | |
| hidden_layer_sizes | [997] | |
| alpha | 8.20e–1 | |
| KNNs | n_neighbors | 11 |
| metric | canberra | |
| weights | uniform | |
| Linear Classifier | learning_rate | invscaling |
| l1_ratio | 7.85e–1 | |
| power_t | 8.51e–2 | |
| loss | modified-huber | |
| penalty | l2 | |
| eta0 | 2.99e–6 | |
| alpha | 4.94e–1 | |
| Random Forest | criterion | entropy |
| max_depth | 100 | |
| min_samples_leaf | 1 | |
| min_weight_fraction_leaf | 3.87e–4 | |
| n_estimators | 401 | |
| max_leaf_nodes | None | |
| min_samples_split | 4 | |
| min_impurity_decrease | 2.89e–4 | |
| Decision Tree | max_depth | 100 |
| min_samples_leaf | 1 | |
| min_weight_fraction_leaf | 2.53e–3 | |
| min_samples_split | 2 | |
| criterion | gini | |
| min_impurity_decrease | 7.73e–5 | |
| max_features | 250 | |
| splitter | best | |
| max_leaf_nodes | None |