如何为keras LSTM处理输入和输出形状

Question

我正在学习RNN，我使用sklearn生成的样本数据集在keras（theano）中编写了这个简单的LSTM模型。

from sklearn.datasets import make_regression
from keras.models import Sequential
from keras.layers import Dense,Activation,LSTM

#creating sample dataset
X,Y=make_regression(100,9,9,2)
X.shape
Y.shape

#creating LSTM model
model = Sequential()
model.add(LSTM(32, input_dim=9))
model.add(Dense(2))
model.compile(loss='mean_squared_error', optimizer='adam')

#model fitting
model.fit(X, Y, nb_epoch=1, batch_size=32)

样本数据集包含9个要素和2个目标。 当我尝试使用这些功能拟合模型并确定目标时，给了我这个错误

Exception: Error when checking model input: expected lstm_input_9 to have 3 dimensions, but got array with shape (100, 9)

Answer 1

如果我是正确的话，那么LSTM需要3D输入。

X = np.random.random((100, 10, 64))
y = np.random.random((100, 2))

model = Sequential()
model.add(LSTM(32, input_shape=(10, 64)))
model.add(Dense(2)) 
model.compile(loss='mean_squared_error', optimizer='adam')

model.fit(X, Y, nb_epoch=1, batch_size=32)

更新 ：如果要将X, Y = make_regression(100, 9, 9, 2)转换为3D，则可以使用它。

from sklearn.datasets import make_regression
from keras.models import Sequential
from keras.layers import Dense,Activation,LSTM

#creating sample dataset
X, Y = make_regression(100, 9, 9, 2)
X = X.reshape(X.shape + (1,))

#creating LSTM model
model = Sequential()
model.add(LSTM(32, input_shape=(9, 1)))
model.add(Dense(2))
model.compile(loss='mean_squared_error', optimizer='adam')

model.fit(X, Y, nb_epoch=1, batch_size=32)