Question

我尝试预测网格图中的2D坐标序列。要做到这一点，我有一个Input_train形状(41185, 10, 2)的训练集，所以我有41185个序列的例子，它们是10个步长，每个步骤加上2个坐标。

这是我的模特：

self.model = Sequential()
                self.model.add(LSTM(self.num_neurons, input_shape=(10, 2), dropout=self.dropout, recurrent_dropout=self.dropout,
                         return_sequences=True))
                for _ in range(self.depth - 1):
                    self.model.add(
                        LSTM(self.num_neurons, dropout=self.dropout, recurrent_dropout=self.dropout,
                             return_sequences=True))
                self.model.add(
                    LSTM(self.num_neurons, dropout=self.dropout, recurrent_dropout=self.dropout))
                self.model.add(Dense(units=3802, activation='relu'))
                self.model.compile(loss=self._get_loss(), optimizer=self._build_optimizer())

我的网络输出应为3802 x 3802阵列，未来坐标在相应的条目中标记为1。

电话：

model.summary()
print "Inputs: {}".format(model.input_shape)
print "Outputs: {}".format(model.output_shape)
print "Actual input: {}".format(Input_train.shape)
print "Actual output: {}".format(Labels_train[0].shape)

产率：

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
lstm_1 (LSTM)                (None, None, 258)         269352    
_________________________________________________________________
lstm_2 (LSTM)                (None, None, 258)         533544    
_________________________________________________________________
lstm_3 (LSTM)                (None, 258)               533544    
_________________________________________________________________
dense_1 (Dense)              (None, 3802)              984718    
=================================================================
Total params: 2,321,158
Trainable params: 2,321,158
Non-trainable params: 0
_________________________________________________________________
Inputs: (None, 10, 2)
Outputs: (None, 3802)
Actual input: (41185, 10, 2)
Actual output: (3802, 3802)

但是在我打电话之后：

history = self.model.fit(Input_train, Labels_train, shuffle=True, validation_data=(Input_test, Labels_test),
                                     verbose=self.verbose)

发生以下错误：

ValueError: Error when checking model target: the list of Numpy arrays that you are passing to your model is not the size the model expected. Expected to see 1 array(s), but instead got the following list of 41185 arrays: [array([[0., 0., 0., ..., 0., 0., 0.],
       [0., 0., 0., ..., 0., 0., 0.],
       [0., 0., 0., ..., 0., 0., 0.],
       ...,
       [0., 0., 0., ..., 0., 0., 0.],
       [0., 0., 0., ..., 0., 0., 0....

我的错误是什么？

Answer 1

将每个2D数组转换为1D，然后尝试将其提供给LSTM，并在检索时将其转换回2D。

如何将2D数据序列馈送到Keras Sequential LSTM

1 个答案: