In the table below, the performance of the optimal hyperparameter setting on the test set for this split is shown, as measured by the optimization metric used for model selection and model assessment. This hyperparameter setting was chosen to be the optimal one in the inner cross-validation (i.e., during the selection) for this assessment split.
| Optimal hyperparameter settings (preprocessing, encoding, ML method) |
Performance (auc) | Reports |
|---|---|---|
| kmer_frequency_logistic_regression | 0.922 | see reports |
In the table below, the performance of the other hyperparameter settings on the test set for this split are shown, as measured by the optimization metric used for model selection and model assessment. These settings were not chosen as the optimal ones during the selection for this assessment split.
| Hyperparameter settings (preprocessing, encoding, ML method) |
Performance (auc) | Reports |
|---|---|---|
| one_hot_cnn | 0.855 | see reports |
| tcrdist_enc_tcrdist_cls | 0.913 | see reports |
The performance of different hyperparameter settings on all listed metrics (both the optimization metric and other metrics not used for model selection and assessment) is shown in the table below. This is the performance on the test set for this assessment split when all models have been retrained on training and validation dataset (the same as the previous table).
| Hyperparameter settings (preprocessing, encoding, ML method) | balanced_accuracy | precision | recall |
| one_hot_cnn | 0.798 | 0.85 | 0.72 |
| tcrdist_enc_tcrdist_cls | 0.756 | 0.696 | 0.896 |
| kmer_frequency_logistic_regression | 0.865 | 0.892 | 0.829 |
For the performance of each of the settings during the inner loop of cross-validation (used to select the optimal model for the split), see selection details.
Here the reports on the datasets before preprocessing and encoding are shown.