tf rnn输入输出的维度和权重的维度

demo

import tensorflow as tf
from tensorflow import keras
from tensorflow.keras import Sequential, Model
from tensorflow.keras.layers import Input, LSTM, Dense

import numpy as np

x = np.random.random(size=(10,4,2))

inputs = Input(shape=(4,2))
lstm1 = LSTM(units=3, return_sequences=False, return_state=False)(inputs)
lstm2 = LSTM(units=3, return_sequences=True, return_state=False)(inputs)
lstm3 = LSTM(units=3, return_sequences=True, return_state=True)(inputs)
lstm4 = LSTM(units=3, return_sequences=False, return_state=True)(inputs)

model1 = Model(inputs=inputs, outputs=lstm1)
model2 = Model(inputs=inputs, outputs=lstm2)
model3 = Model(inputs=inputs, outputs=lstm3)
model4 = Model(inputs=inputs, outputs=lstm4)
res1 = model1.predict(x)
res2 = model2.predict(x)
res3 = model3.predict(x)
res4 = model4.predict(x)

print("res1只有一个返回值")
print(res1.shape)

print("res2只有一个返回值")
print(res2.shape)

print("res3有三个返回值")
for i in res3:
	print(i.shape)

上面的代码表示一个
$x_i[10,2]\ast w_x[2,3]=[10,3]$
$h_i[10,3]\ast w_h[3,3]=[10,3]$

return参数

lstm中有两个参数
return_sequences
return_state

return_sequences

  return_sequences: Boolean. Whether to return the last output in the output sequence, or the full sequence. Default: `False`.

这里是选择是否返回last_output，那么什么是last_output
output就是每个cell的output总和，每一个cell的输出为$[10,3]$，所有的output为$[10,4,3]$
last_output就是所有cell总和中最后一个输出项

return_state

return_state: Boolean. Whether to return the last state in addition to the output. Default: `False`

state就是$h$，它的维度是$[10,3]$

ref

参考一