Question

我正在尝试在Keras中实施多步预测LSTM模型。数据的形状是这样的：

X : (5831, 48, 1)
y : (5831, 1, 12)

我要使用的模型是：

power_in = Input(shape=(X.shape[1], X.shape[2]))


power_lstm = LSTM(50, recurrent_dropout=0.4128,
                  dropout=0.412563, kernel_initializer=power_lstm_init, return_sequences=True)(power_in)

main_out = TimeDistributed(Dense(12, kernel_initializer=power_lstm_init))(power_lstm)

在尝试像这样训练模型时：

hist = forecaster.fit([X], y, epochs=325, batch_size=16, validation_data=([X_valid], y_valid), verbose=1, shuffle=False)

我遇到以下错误：

ValueError: Error when checking target: expected time_distributed_16 to have shape (48, 12) but got array with shape (1, 12)

如何解决此问题？

Answer 1

根据您的评论：

我拥有的数据像t-48，t-47，t-46，.....，t-1一样，是过去的数据， t + 1，t + 2，......，t + 12作为我要预测的值

您可能根本不需要使用TimeDistributed层：首先，只需删除LSTM层的resturn_sequences=True参数。完成此操作后，LSTM层将以形状为(50,)的向量对过去的输入时间序列进行编码。现在，您可以将其直接喂入具有12个单位的密集层：

# make sure the labels have are in shape (num_samples, 12)
y = np.reshape(y, (-1, 12))

power_in = Input(shape=(X.shape[1:],))
power_lstm = LSTM(50, recurrent_dropout=0.4128,
                  dropout=0.412563,
                  kernel_initializer=power_lstm_init)(power_in)

main_out = Dense(12, kernel_initializer=power_lstm_init)(power_lstm)

或者，如果您想使用TimeDistributed层并考虑到输出本身是序列，则可以在模型中使用Dense层之前的另一个LSTM层（使用在第一个LSTM层之后添加一个RepeatVector层，以使其输出的长度为12的时间序列，即与输出时间序列的长度相同）：

# make sure the labels have are in shape (num_samples, 12, 1)
y = np.reshape(y, (-1, 12, 1))

power_in = Input(shape=(48,1))
power_lstm = LSTM(50, recurrent_dropout=0.4128,
                  dropout=0.412563,
                  kernel_initializer=power_lstm_init)(power_in)

rep = RepeatVector(12)(power_lstm)
out_lstm = LSTM(32, return_sequences=True)(rep)
main_out = TimeDistributed(Dense(1))(out_lstm)

model = Model(power_in, main_out)
model.summary()

模型摘要：

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
input_3 (InputLayer)         (None, 48, 1)             0         
_________________________________________________________________
lstm_3 (LSTM)                (None, 50)                10400     
_________________________________________________________________
repeat_vector_2 (RepeatVecto (None, 12, 50)            0         
_________________________________________________________________
lstm_4 (LSTM)                (None, 12, 32)            10624     
_________________________________________________________________
time_distributed_1 (TimeDist (None, 12, 1)             33        
=================================================================
Total params: 21,057
Trainable params: 21,057
Non-trainable params: 0
_________________________________________________________________

当然，在两个模型中，您可能都需要调整超参数（例如LSTM层的数量，LSTM层的尺寸等），以便能够准确地比较它们并获得良好的结果。

旁注：：实际上，在您的情况下，您根本不需要使用TimeDistributed层，因为（当前）Dense layer is applied on the last axis。因此，TimeDistributed(Dense(...))和Dense(...)是等效的。

多步预测LSTM模型

1 个答案: