comparison optimise_hyperparameters.py @ 2:76251d1ccdcc draft

"planemo upload for repository https://github.com/bgruening/galaxytools/tree/recommendation_training/tools/tool_recommendation_model commit 6fa2a0294d615c9f267b766337dca0b2d3637219"

1:12764915e1c5

    @classmethod
    def __init__(self):
        """ Init method. """

    @classmethod
    def train_model(self, config, reverse_dictionary, train_data, train_labels, test_data, test_labels, class_weights):
        """
        Train a model and report accuracy
        """
        l_recurrent_activations = config["activation_recurrent"].split(",")
        l_output_activations = config["activation_output"].split(",")

        validation_split = float(config["validation_share"])

        # get dimensions
        dimensions = len(reverse_dictionary) + 1
        best_model_params = dict()
        early_stopping = EarlyStopping(monitor='val_loss', mode='min', min_delta=1e-4, verbose=1, patience=1)

        # specify the search space for finding the best combination of parameters using Bayesian optimisation
        params = {
            "embedding_size": hp.quniform("embedding_size", l_embedding_size[0], l_embedding_size[1], 1),
            "units": hp.quniform("units", l_units[0], l_units[1], 1),

                shuffle="batch",
                verbose=2,
                validation_split=validation_split,
                callbacks=[early_stopping]
            )
            return {'loss': model_fit.history["val_loss"][-1], 'status': STATUS_OK}
        # minimize the objective function using the set of parameters above4
        trials = Trials()
        learned_params = fmin(create_model, params, trials=trials, algo=tpe.suggest, max_evals=int(config["max_evals"]))
        print(learned_params)
        # set the best params with respective values
        for item in learned_params:
            item_val = learned_params[item]
            if item == 'activation_output':
                best_model_params[item] = l_output_activations[item_val]
            elif item == 'activation_recurrent':
                best_model_params[item] = l_recurrent_activations[item_val]
            else:
                best_model_params[item] = item_val
        return best_model_params

2:76251d1ccdcc

    @classmethod
    def __init__(self):
        """ Init method. """

    @classmethod
    def train_model(self, config, reverse_dictionary, train_data, train_labels, class_weights):
        """
        Train a model and report accuracy
        """
        l_recurrent_activations = config["activation_recurrent"].split(",")
        l_output_activations = config["activation_output"].split(",")

        validation_split = float(config["validation_share"])

        # get dimensions
        dimensions = len(reverse_dictionary) + 1
        best_model_params = dict()
        early_stopping = EarlyStopping(monitor='val_loss', mode='min', verbose=1, min_delta=1e-4)

        # specify the search space for finding the best combination of parameters using Bayesian optimisation
        params = {
            "embedding_size": hp.quniform("embedding_size", l_embedding_size[0], l_embedding_size[1], 1),
            "units": hp.quniform("units", l_units[0], l_units[1], 1),

                shuffle="batch",
                verbose=2,
                validation_split=validation_split,
                callbacks=[early_stopping]
            )
            return {'loss': model_fit.history["val_loss"][-1], 'status': STATUS_OK, 'model': model}
        # minimize the objective function using the set of parameters above
        trials = Trials()
        learned_params = fmin(create_model, params, trials=trials, algo=tpe.suggest, max_evals=int(config["max_evals"]))
        best_model = trials.results[np.argmin([r['loss'] for r in trials.results])]['model']

        # set the best params with respective values
        for item in learned_params:
            item_val = learned_params[item]
            if item == 'activation_output':
                best_model_params[item] = l_output_activations[item_val]
            elif item == 'activation_recurrent':
                best_model_params[item] = l_recurrent_activations[item_val]
            else:
                best_model_params[item] = item_val
        return best_model_params, best_model