The Grid_Search class provides a simple way to execute a hyperparameter tuning for the convolutional neural network model. Have a look at the Model documentation for an overview of all available hyperparameters. The tuning returns the best model (highest ROC-AUC or PR-AUC on the validation data) and an overview of all trained models.
| name | description |
|---|---|
| __init__ | Initialize the object with a collection of parameter values. |
| train | Train all models and return the best one. |
def __init__(self, params)Initialize the object with a collection of parameter values.
For example: providing {'conv_num': [1,2,3], 'kernel_num': [20,50]} will result in training 6 different models (all possible combinations of the provided values) when the train() method is called later on. Parameters that are not provided here will hold their default values in all 6 models.
| parameter | type | description |
|---|---|---|
| params | dict | A dict containing parameter names as keys and corresponding values as lists. |
def train(self, data, pr_auc = False, verbose = True)Train all models and return the best one.
Models are evaluated and ranked according to their ROC-AUC or PR-AUC (precision-recall) on a validation data set.
| parameter | type | description |
|---|---|---|
| data | pysster.Data | A Data object providing training and validation data sets. |
| pr_auc | bool | If True, the area under the precision-recall curve will be maximized instead of the area under the ROC curve |
| verbose | bool | If True, progress information (train/val loss) will be printed throughout the training. |
| returns | type | description |
|---|---|---|
| results | tuple(pysster.Model, str) | The best performing model and an overview table of all models are returned. |