scripts.nn#

Functions#

`build_model`(input_dim, num_classes)	Helper function to build a fresh model.
`calculate_threshold`(uncertainties, y_pred_classes, ...)	Returns the threshold that gives the best accuracy, not a quantile but the actual threshold.
`predict_validation`(model, X_val_bf, y_val_bf, ...)	Predicts the species of the test data.
`prepare_for_training`([TrainPanel])
`save_train_stats`(model, X_val_bf, y_val_bf, ...[, ...])
`train_neural_network`([TrainPanel])
`train_wrapper`(model, X_train, y_train, y_train_int, ...)	Helper function to train the model and return validation accuracy.

Module Contents#

scripts.nn.build_model(input_dim, num_classes)#: Helper function to build a fresh model.

scripts.nn.calculate_threshold(uncertainties, y_pred_classes, y_true_classes, species_names)#: Returns the threshold that gives the best accuracy, not a quantile but the actual threshold. Thus, its value can be higher than 1.0.

scripts.nn.predict_validation(model, X_val_bf, y_val_bf, species_names)#: Predicts the species of the test data.

scripts.nn.prepare_for_training(TrainPanel=None, **kwargs)#

scripts.nn.save_train_stats(model, X_val_bf, y_val_bf, species_names, model_dir, best_accuracy, fold_count, best_fold=None)#

scripts.nn.train_neural_network(TrainPanel=None, **kwargs)#

scripts.nn.train_wrapper(model, X_train, y_train, y_train_int, X_val, y_val, epochs, batch_size, patience)#: Helper function to train the model and return validation accuracy.