了解 Keras LSTM Tensorboard 图

Question

我对在 Tensborboad 中为我的 Keras LSTM 网络获取的图表感到困惑。我已经这样定义了我的 Keras LSTM 网络：

model = Sequential()
model.add(LSTM(neurons, return_sequences=True,input_shape=(look_back,2)))
#model.add(Bidirectional(LSTM(neurons, return_sequences=True),input_shape=(look_back,2)))
model.add(LSTM(neurons,return_sequences=True,recurrent_regularizer=l2(weight_decay),
          kernel_regularizer=l2(weight_decay),bias_regularizer=l2(weight_decay),dropout=dropout,recurrent_dropout=dropout))
model.add(LSTM(neurons,return_sequences=True,recurrent_regularizer=l2(weight_decay),
          kernel_regularizer=l2(weight_decay),bias_regularizer=l2(weight_decay),dropout=dropout,recurrent_dropout=dropout))
model.add(LSTM(neurons,return_sequences=True,recurrent_regularizer=l2(weight_decay),
          kernel_regularizer=l2(weight_decay),bias_regularizer=l2(weight_decay),dropout=dropout,recurrent_dropout=dropout))
model.add(LSTM(neurons,return_sequences=True,recurrent_regularizer=l2(weight_decay),
          kernel_regularizer=l2(weight_decay),bias_regularizer=l2(weight_decay),dropout=dropout,recurrent_dropout=dropout))
model.add(LSTM(neurons,return_sequences=True,recurrent_regularizer=l2(weight_decay),
          kernel_regularizer=l2(weight_decay),bias_regularizer=l2(weight_decay),dropout=dropout,recurrent_dropout=dropout))
model.add(LSTM(neurons,return_sequences=True,recurrent_regularizer=l2(weight_decay),
          kernel_regularizer=l2(weight_decay),bias_regularizer=l2(weight_decay),dropout=dropout,recurrent_dropout=dropout))
model.add(LSTM(neurons,return_sequences=True,recurrent_regularizer=l2(weight_decay),
          kernel_regularizer=l2(weight_decay),bias_regularizer=l2(weight_decay),dropout=dropout,recurrent_dropout=dropout))
model.add(LSTM(neurons,return_sequences=True,recurrent_regularizer=l2(weight_decay),
          kernel_regularizer=l2(weight_decay),bias_regularizer=l2(weight_decay),dropout=dropout,recurrent_dropout=dropout))
model.add(LSTM(neurons,return_sequences=True,recurrent_regularizer=l2(weight_decay),
          kernel_regularizer=l2(weight_decay),bias_regularizer=l2(weight_decay),dropout=dropout,recurrent_dropout=dropout))
model.add(LSTM(neurons,return_sequences=True,recurrent_regularizer=l2(weight_decay),
          kernel_regularizer=l2(weight_decay),bias_regularizer=l2(weight_decay),dropout=dropout,recurrent_dropout=dropout))
model.add(LSTM(neurons,return_sequences=True,recurrent_regularizer=l2(weight_decay),
          kernel_regularizer=l2(weight_decay),bias_regularizer=l2(weight_decay),dropout=dropout,recurrent_dropout=dropout))   
model.add(LSTM(20,return_sequences=False,recurrent_regularizer=l2(weight_decay),
          kernel_regularizer=l2(weight_decay),bias_regularizer=l2(weight_decay),dropout=dropout,recurrent_dropout=dropout))
model.add(Dense(outputs,kernel_regularizer=l2(weight_decay),bias_regularizer=l2(weight_decay),activation='linear'))
model.compile(loss='mean_squared_error', optimizer='adam', metrics=['accuracy'])

我认为这会给我一个顺序模型，其中每个 LSTM 都采用前一个 LSTM 的输出。我有点明白了。但是我也得到了一个 LSTM 层作为每个后续层的输入：

在图表中，看起来 lstm_2 馈入每一层。我不会预料到的。所以我的问题是，这是预期的吗？如果是这样，为什么？

谢谢。

Answer 1

我明白为什么会这样显示了。事实证明，Keras 创建了一个 learning_phase 占位符并将其放置在第二个隐藏层中。 learning_phase 对象分支到每一层，但 LSTM 本身没有。我参考了解更多详情。

这是我的 LSTM_1 层的内部在我的 Tensorboard 图表中的样子：

了解 Keras LSTM Tensorboard 图

Understanding Keras LSTM Tensorboard Graph

python

tensorflow

keras

lstm

tensorboard