add random seed setting (fixed #9)

            est_hiddens, est_activation, est_dropout_ratio=0.5,

            minibatch_size=1024, epoch_size=100,

            learning_rate=0.0001, lambda1=0.1, lambda2=0.005,

            normalize=True):

            normalize=True, random_seed=123):

        """

        Parameters

        ----------

        normalize : bool (optional)

            specify whether input data need to be normalized.

            by default, input data is normalized.

        random_seed : int (optional)

            random seed used when fit() is called.

        """

        self.comp_net = CompressionNet(comp_hiddens, comp_activation)

        self.est_net = EstimationNet(est_hiddens, est_activation)

        self.normalize = normalize

        self.scaler = None

        self.seed = random_seed

        self.graph = None

        self.sess = None

        with tf.Graph().as_default() as graph:

            self.graph = graph

            tf.set_random_seed(self.seed)

            # Create Placeholder

            self.input = input = tf.placeholder(