张量流中两个RNN实现之间有什么区别?

时间:2016-05-16 11:35:40

标签: python tensorflow deep-learning lstm recurrent-neural-network

我在tensorflow中找到了两种RNN实现。

第一个实现是this(从第124行到第129行)。它使用循环来定义RNN中输入的每个步骤。

with tf.variable_scope("RNN"):
      for time_step in range(num_steps):
        if time_step > 0: tf.get_variable_scope().reuse_variables()
        (cell_output, state) = cell(inputs[:, time_step, :], state)
        outputs.append(cell_output)
        states.append(state)

第二个实现是this(从第51行到第70行)。它不使用任何循环来定义RNN中的每个输入步骤。

def RNN(_X, _istate, _weights, _biases):

    # input shape: (batch_size, n_steps, n_input)
    _X = tf.transpose(_X, [1, 0, 2])  # permute n_steps and batch_size
    # Reshape to prepare input to hidden activation
    _X = tf.reshape(_X, [-1, n_input]) # (n_steps*batch_size, n_input)
    # Linear activation
    _X = tf.matmul(_X, _weights['hidden']) + _biases['hidden']

    # Define a lstm cell with tensorflow
    lstm_cell = rnn_cell.BasicLSTMCell(n_hidden, forget_bias=1.0)
    # Split data because rnn cell needs a list of inputs for the RNN inner loop
    _X = tf.split(0, n_steps, _X) # n_steps * (batch_size, n_hidden)

    # Get lstm cell output
    outputs, states = rnn.rnn(lstm_cell, _X, initial_state=_istate)

    # Linear activation
    # Get inner loop last output
    return tf.matmul(outputs[-1], _weights['out']) + _biases['out']

在第一个实现中,我发现输入单元与隐藏单元之间没有权重矩阵,只定义隐藏单元到输出单元之间的权重矩阵(从132到133行).. < / p>

output = tf.reshape(tf.concat(1, outputs), [-1, size])
        softmax_w = tf.get_variable("softmax_w", [size, vocab_size])
        softmax_b = tf.get_variable("softmax_b", [vocab_size])
        logits = tf.matmul(output, softmax_w) + softmax_b

但在第二个实现中,两个权重矩阵都已定义(从第42行到第47行)。

weights = {
    'hidden': tf.Variable(tf.random_normal([n_input, n_hidden])), # Hidden layer weights
    'out': tf.Variable(tf.random_normal([n_hidden, n_classes]))
}
biases = {
    'hidden': tf.Variable(tf.random_normal([n_hidden])),
    'out': tf.Variable(tf.random_normal([n_classes]))
}

我想知道为什么?

1 个答案:

答案 0 :(得分:3)

我注意到的差异是second implementation中的代码使用tf.nn.rnn,它获取每个时间步的输入列表并生成每个时间步的输出列表。

(输入:输入的长度T列表,每个都是一个形状的张量       [batch_size,input_size]。)

因此,如果你检查第62行第二个实现中的代码,输入数据将被整形为n_steps *(batch_size,n_hidden)

# Split data because rnn cell needs a list of inputs for the RNN inner loop
_X = tf.split(0, n_steps, _X) # n_steps * (batch_size, n_hidden)

1st implementation中,他们循环遍历n_time_steps并提供输入并获取相应的输出并存储在输出列表中。

第113到117行的代码段

outputs = []
    state = self._initial_state
    with tf.variable_scope("RNN"):
      for time_step in range(num_steps):
        if time_step > 0: tf.get_variable_scope().reuse_variables()
        (cell_output, state) = cell(inputs[:, time_step, :], state)
        outputs.append(cell_output)

来到你的第二个问题:

如果您在两个实现中都仔细注意了输入被输入RNN的方式。

在第一个实现中,输入的形状已经是batch_size x num_steps(这里num_steps是隐藏的大小):

self._input_data = tf.placeholder(tf.int32, [batch_size, num_steps])

而在第二种实现中,初始输入具有形状(batch_size x n_steps x n_input)。因此需要一个权重矩阵来转换为形状(n_steps x batch_size x hidden_​​size):

    # Input shape: (batch_size, n_steps, n_input)
    _X = tf.transpose(_X, [1, 0, 2])  # Permute n_steps and batch_size
    # Reshape to prepare input to hidden activation
    _X = tf.reshape(_X, [-1, n_input]) # (n_steps*batch_size, n_input)
    # Linear activation
    _X = tf.matmul(_X, _weights['hidden']) + _biases['hidden']
    # Split data because rnn cell needs a list of inputs for the RNN inner loop
    _X = tf.split(0, n_steps, _X) # n_steps * (batch_size, n_hidden)

我希望这有用......