Question

我强烈怀疑有更好的方法来设置它，欢迎提出建议，但我在这里：

我想在Keras中使用递归神经网络对时间序列数据进行多类分类。我的模型定义如下：

model = Sequential()
model.add(LSTM(out_dim, input_shape = (X_train.shape[1], X_train.shape[2]), return_sequences = False))
model.add(Dense(num_classes, activation='sigmoid'))

optim_type = ["rmsprop", "adam", "sgd"]
  for optim_val in optim_type:
            if optim_val == "sgd" and default_val == False:
                   ...
            else:
                    optim_use = optim_type
            model.compile(loss='categorical_crossentropy', optimizer = optim_use, metrics = ['accuracy'])
            hist = model.fit(X_train, dummy_y, validation_data=(X_test, dummy_y_test), nb_epoch = epochs, batch_size = b_size)

我得到的错误是：

Using TensorFlow backend.
Traceback (most recent call last):
  File "../rnn_new.py", line 213, in <module>
    network_LSTM_rnn(data_in, out_dim, optim_type, b_size, save_file, num_classes, epochs, default_val)
  File "../rnn_new.py", line 166, in network_LSTM_rnn
    hist = model.fit(X_train, dummy_y, validation_data=(X_test, dummy_y_test), nb_epoch = epochs, batch_size = b_size)
  File "/user/pkgs/anaconda2/envs/my_env/lib/python2.7/site-packages/keras/models.py", line 627, in fit
    sample_weight=sample_weight)
  File "/user/pkgs/anaconda2/envs/my_env/lib/python2.7/site-packages/keras/engine/training.py", line 1097, in fit
    self._make_train_function()
  File "/user/pkgs/anaconda2/envs/my_env/lib/python2.7/site-packages/keras/engine/training.py", line 712, in _make_train_function
    training_updates = self.optimizer.get_updates(self._collected_trainable_weights,
AttributeError: 'list' object has no attribute 'get_updates'

如何修复此错误？你还需要我发布什么吗？

Answer 1

我只是无法命名变量。我有optim_use = optim_type我应该有optim_use = optim_val。不管怎样，谢谢你！

Keras：循环中的模型定义：'list'对象没有属性

1 个答案: