CNN ValueError 当我尝试拟合我的模型时

Question

这是我的代码：

from keras import optimizers from keras.layers import Convolution1D, Dense, MaxPooling1D, Flatten
model = Sequential() 
# my CNN layers
model.add(Conv1D(101, 101, strides=1, padding='same', dilation_rate=1, input_shape=(None, 120)))
model.add(Activation('relu')) model.add(MaxPooling1D(pool_size=2, padding='same', strides=None))
model.add(Dense(2048)) model.add(Activation('relu'))
model.add(Dense(100)) model.add(Activation('sigmoid'))
model.compile(optimizer=optimizers.Adam(lr=1e-4), loss='binary_crossentropy', metrics=['accuracy'])
model.fit(training_trainX_train, training_trainY_train, epochs=2, batch_size=100, verbose=1)

但是我收到此错误：ValueError: Error when checking model input: expected conv1d_8_input to have 3 dimensions, but got array with shape (27660, 120)

这是我的训练集的形状：

training_trainX_train.shape = (27660, 120) training_trainY_train.shape = (27660, 101)

Answer 1

添加model.add(Flatten())将解决此问题

model.add(Conv1D(101, 101, strides=1, padding='same', dilation_rate=1,       input_shape=(None, 120)))
model.add(Activation('relu')) model.add(MaxPooling1D(pool_size=2, padding='same', strides=None))
model.add(Flatten())
model.add(Dense(2048)) model.add(Activation('relu'))
model.add(Dense(100)) model.add(Activation('sigmoid'))

有关更多详细信息，请查看 https://github.com/keras-team/keras/issues/6351