更新时间:2023-12-02 08:11:46
您完全不需要重组任何结构即可将Conv1D层的输出转换为LSTM层.
You don't need to restructure anything at all to get the output of a Conv1D layer into an LSTM layer.
因此,问题仅在于Flatten
层的存在,该层破坏了形状.
So, the problem is simply the presence of the Flatten
layer, which destroys the shape.
这些是Conv1D和LSTM使用的形状:
These are the shapes used by Conv1D and LSTM:
(batch, length, channels)
(batch, timeSteps, features)
(batch, length, channels)
(batch, timeSteps, features)
长度与timeSteps相同,通道与要素相同.
Length is the same as timeSteps, and channels is the same as features.
使用Bidirectional
包装器也不会改变任何事情.它只会复制您的输出功能.
Using the Bidirectional
wrapper won't change a thing either. It will only duplicate your output features.
如果要对整个序列进行整体分类,则最后一个LSTM必须使用return_sequences=False
. (或者您也可以在之后使用扁平+密集)
If you're going to classify the entire sequence as a whole, your last LSTM must use return_sequences=False
. (Or you may use some flatten + dense instead after)
如果要对序列的每个步骤进行分类,则所有LSTM都应具有return_sequences=True
.您不应该在它们之后将数据弄平.
If you're going to classify each step of the sequence, all your LSTMs should have return_sequences=True
. You should not flatten the data after them.