Table 2
Classifiers considered for ML-based comparisons according to selected parameters.
| ID | Type | # Estimator | Max. depth | # Layers | Units per layer | AUC single pop. | AUC Transformer |
|---|---|---|---|---|---|---|---|
| LC | Linear | 0.57 | 0.54 | ||||
| SVC | Support Vector Classifier | 0.55 | 0.54 | ||||
| RF1 | Random Forrest | 10 | 5 | 0.57 | 0.55 | ||
| RF2 | Random Forrest | 10 | 20 | 0.53 | 0.55 | ||
| RF3 | Random Forrest | 50 | 5 | 0.55 | 0.54 | ||
| RF4 | Random Forrest | 50 | 20 | 0.53 | 0.57 | ||
| DNN1 | Deep Neural Network | 1 | 16 | 0.58 | 0.57 | ||
| DNN2 | Deep Neural Network | 1 | 32 | 0.59 | 0.58 | ||
| DNN3 | Deep Neural Network | 1 | 64 | 0.59 | 0.52 | ||
| DNN4 | Deep Neural Network | 1 | 128 | 0.60 | 0.56 | ||
| DNN5 | Deep Neural Network | 3 | 16 | 0.60 | 0.57 | ||
| DNN6 | Deep Neural Network | 3 | 32 | 0.62 | 0.58 | ||
| DNN7 | Deep Neural Network | 3 | 64 | 0.59 | 0.58 | ||
| DNN8 | Deep Neural Network | 3 | 128 | 0.58 | 0.57 | ||
| DNN9 | Deep Neural Network | 5 | 16 | 0.60 | 0.58 | ||
| DNN10 | Deep Neural Network | 5 | 32 | 0.58 | 0.56 | ||
| DNN11 | Deep Neural Network | 5 | 64 | 0.59 | 0.59 | ||
| DNN12 | Deep Neural Network | 5 | 128 | 0.59 | 0.57 |
Notes. The AUC is obtained when classifying two similar parts of the same population (see text for details). All classifiers were computed using the Sci-kit library (see https://scikit-learn.org/stable/) or (for deep neural networks – DNN) the Keras library (see https://keras.io). The classifier, AUC Single pop., refers to the AUC when classifying the numerical population only (with random labels), while AUC Transformer refers to the AUC for the last iteration of the generative model (baseline model).
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.