env.py 文件源码

def create_training_test_sets(self):
        """ Split data set into training and test folds. """
        # load input data
        input_data = np.asarray(np.loadtxt('input/data.txt'), dtype=np.float32)
        self.input_dim = input_data.shape[1] - 1
        self.output_dim = 1

        # align to batch size
        batches = input_data.shape[0] // (self.batch_size * self.n_splits)
        input_data = input_data[:batches * (self.batch_size * self.n_splits)]

        self.data_size = input_data.shape[0]
        print(f'Loaded input data, shape = {input_data.shape}')

        # create splits
        kfold = KFold(n_splits=self.n_splits, shuffle=True, random_state=self.seed)
        print(f'Splits: {self.n_splits}')

        # assume y is in the last column by default
        for idx_train, idx_test in kfold.split(input_data):
            self.train_x.append(input_data[idx_train, :-1])
            self.train_y.append(input_data[idx_train, -1:])
            self.test_x.append(input_data[idx_test, :-1])
            self.test_y.append(input_data[idx_test, -1:])

        # layers described as [number of neurons, dropout probability]
        if self.layers_description is None:
            self.layers_description = [[self.input_dim, 0.0], [100, 0.0], [100, 0.0], [self.output_dim, 0.0]]