LSTM Wrapper with Gaussian Noise added to Input

gaussian_noise_wrapper.py

class GaussianNoiseWrapper(tf.contrib.rnn.RNNCell):
    """
    Wrapper for RNNCells to add gaussian noise to the input with specified mean,
    stdv before each call of the internal RNNCell.

    (The structure of this wrapper class follows the design of wrappers in tf.contrib.rnn
    such as InputProjectionWrapper, DropoutWrapper, OutputProjectionWrapper, etc.)
    """
    def __init__(self, cell, mean=0.0, stddev=0.1):
        """
        Create a cell with gaussian noise appended. This just stores the mean and standard
        deviation along with the cell.
        """
        if not isinstance(cell, tf.contrib.rnn.RNNCell):
            raise TypeError("The parameter cell is not RNNCell.")
        self._cell = cell
        self._mu = mean
        self._sigma = stddev

    @property
    def state_size(self):
        return self._cell.state_size

    @property
    def output_size(self):
        return self._cell.output_size

    def zero_state(self, batch_size, dtype):
        with tf.name_scope(type(self).__name__ + "ZeroState", values=[batch_size]):
            return self._cell.zero_state(batch_size, dtype)

    def __call__(self, inputs, state, scope=None):
        """Run the input projection and then the cell."""
        # Default scope: "GaussianNoiseWrapper"
        with tf.variable_scope(scope or "gaussian_noise_wrapper"):
            _noise = tf.random_uniform(tf.shape(inputs), mean=self._mu, stddev=self._sigma)
            noise_added = tf.add(inputs, _noise)

            return self._cell(noise_added, state)