Python Keras LSTM 输入输出形状问题

Question

我在 tensorflow 上运行ning keras，试图实现一个多维 LSTM 网络来预测一个线性连续目标变量，每个示例都有一个值（return_sequences = False）。我的序列长度是 10，特征数（暗）是 11。这就是我运行:

import pprint, pickle
import numpy as np
from keras.models import Sequential
from keras.layers import Dense, Activation
from keras.layers import LSTM

# Input sequence
wholeSequence = [[0,0,0,0,0,0,0,0,0,2,1],
                 [0,0,0,0,0,0,0,0,2,1,0],
                 [0,0,0,0,0,0,0,2,1,0,0],
                 [0,0,0,0,0,0,2,1,0,0,0],
                 [0,0,0,0,0,2,1,0,0,0,0],
                 [0,0,0,0,2,1,0,0,0,0,0],
                 [0,0,0,2,1,0,0,0,0,0,0],
                 [0,0,2,1,0,0,0,0,0,0,0],
                 [0,2,1,0,0,0,0,0,0,0,0],
                 [2,1,0,0,0,0,0,0,0,0,0]]

# Preprocess Data:
wholeSequence = np.array(wholeSequence, dtype=float) # Convert to NP array.
data = wholeSequence
target = np.array([20])

# Reshape training data for Keras LSTM model
data = data.reshape(1, 10, 11)
target = target.reshape(1, 1, 1)

# Build Model
model = Sequential()
model.add(LSTM(11, input_shape=(10, 11), unroll=True, return_sequences=False))
model.add(Dense(11))
model.add(Activation('linear'))
model.compile(loss='mean_squared_error', optimizer='adam')
model.fit(data, target, nb_epoch=1, batch_size=1, verbose=2)

并得到错误 ValueError: Error when checking target: expected activation_1 to have 2 dimensions, but got array with shape (1, 1, 1) 不确定激活层应该得到什么（形状方面）任何帮助表示赞赏谢谢

Answer 1

如果你只想有一个线性输出神经元，你可以简单地使用一个带有一个隐藏单元的密集层并在那里提供激活。然后你的输出可以是没有整形的单个向量-我调整了你给定的示例代码以使其工作：

wholeSequence = np.array(wholeSequence, dtype=float) # Convert to NP array.
data = wholeSequence
target = np.array([20])

# Reshape training data for Keras LSTM model
data = data.reshape(1, 10, 11)

# Build Model
model = Sequential()
model.add(LSTM(11, input_shape=(10, 11), unroll=True, return_sequences=False))
model.add(Dense(1, activation='linear'))
model.compile(loss='mean_squared_error', optimizer='adam')
model.fit(data, target, nb_epoch=1, batch_size=1, verbose=2)

Python Keras LSTM 输入输出形状问题

Python Keras LSTM input output shape issue

python-3.x

tensorflow

keras

shape

activation-function