LSTM参数个数计算方法

计算LSTM的参数个数需要了解其工作原理，如下：

可以看到若不加偏置，该LSTM层共有4个权重参数，图中输入的一个timestep的特征大小为3（绿色单元），隐藏层输出大小为2（橙色单元），因此该LSTM层的参数个数为：
$n=4*(2+3)*2=40$

若加上偏置则该LSTM层的参数个数为：
$n=4*((2+3)*2+2)=48$

使用keras的LSTM进行验证

带偏置：

from keras.models import Sequential
from keras.layers import LSTM

timestep = 4  # LSTM层的block个数
model = Sequential()
model.add(LSTM(units=2, use_bias=True, input_shape=[timestep, 3]))
model.summary()

Model: "sequential_1"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
lstm_1 (LSTM)                (None, 2)                 48        
=================================================================
Total params: 48
Trainable params: 48
Non-trainable params: 0
_________________________________________________________________

不带偏置：

from keras.models import Sequential
from keras.layers import LSTM

timestep = 4  # LSTM层的block个数
model = Sequential()
model.add(LSTM(units=2, use_bias=False, input_shape=[timestep, 3]))
model.summary()

Model: "sequential_1"
_________________________________________________________________
Layer (type)                 Output Shape              Param #
=================================================================
lstm_1 (LSTM)                (None, 2)                 40
=================================================================
Total params: 40
Trainable params: 40
Non-trainable params: 0
_________________________________________________________________

以小见大，通过推理可以得到一般的计算方法。