ValueError：检查输入时出错：预期 conv2d_1_input 有 4 个维度，但在与 model 拟合时得到形状为 (999, 12, 1) 的数组

Question

我正在我的数据集上构建一个二维卷积网络。我在一个测试集上运行它，代码如下：

#reproducible code
from keras.models import Sequential
from keras.layers import Dense, Dropout, Flatten
from keras.utils import np_utils
from keras import optimizers
from sklearn.metrics import confusion_matrix
import numpy as np
import time
from keras.layers.convolutional import Conv2D

data = np.random.rand(1000,22)
data.shape
train_X = data[0:data.shape[0],0:12]
train_X.shape
train_y = data[0:data.shape[0],12:data.shape[1]]
train_y.shape
train_X = train_X.reshape((train_X.shape[0], train_X.shape[1], 1))
train_X.shape
neurons = 10
model = Sequential()
model.add(Conv2D(filters=64,input_shape=train_X.shape, activation='relu',kernel_size = 3))
model.add(Flatten())
model.add(Dense(neurons,activation='relu')) # first hidden layer
model.add(Dense(neurons, activation='relu')) # second hidden layer
model.add(Dense(neurons, activation='relu')) # third hidden layer
model.add(Dense(10, activation='softmax'))
sgd = optimizers.SGD(lr=0.05, decay=1e-6, momentum=0.95, nesterov=True)
model.compile(loss='categorical_crossentropy', optimizer=sgd, metrics=['accuracy'])
model.summary()
model.fit(train_X,train_y, validation_split=0.2, epochs=10, batch_size=100, verbose=0)
model.summary()

我的 model 运行了一段时间并显示以下摘要：

Model: "sequential_1"
_________________________________________________________________
Layer (type)                 Output Shape              Param #
================================================================= 
conv2d_1 (Conv2D)            (None, 997, 10, 64)       640       
_________________________________________________________________
flatten_1 (Flatten)          (None, 638080)            0
_________________________________________________________________
dense_1 (Dense)              (None, 10)                6380810
_________________________________________________________________
dense_2 (Dense)              (None, 10)                110
_________________________________________________________________
dense_3 (Dense)              (None, 10)                110
_________________________________________________________________
dense_4 (Dense)              (None, 10)                110
=================================================================
Total params: 6,381,780
Trainable params: 6,381,780
Non-trainable params: 0

它卡在 model.fit 并抛出如下所示的错误。我想知道如何解决这个错误。

Traceback (most recent call last):
  File "CNN_test.py", line 65, in <module>
    model.fit(train_X,train_y, validation_split=0.2, epochs=10, batch_size=100, verbose=0)
  File "/usr/local/lib/python3.6/site-packages/keras/engine/training.py", line 1154, in fit
    batch_size=batch_size)
  File "/usr/local/lib/python3.6/site-packages/keras/engine/training.py", line 579, in _standardize_user_data
    exception_prefix='input')
  File "/usr/local/lib/python3.6/site-packages/keras/engine/training_utils.py", line 135, in standardize_input_data
    'with shape ' + str(data_shape))
ValueError: Error when checking input: expected conv2d_1_input to have 4 dimensions, but got array with shape (999, 12, 1)

Answer 1

您没有理由使用 2D 卷积层，因为您的数据是 3D。 您正在寻找的是 Conv1D。 此外，不要在input_shape中包含n_samples维度。

from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense, Conv1D, Flatten
from tensorflow.keras import optimizers
import numpy as np

data = np.random.rand(1000,22)

train_X = data[0:data.shape[0],0:12]
train_X = train_X.reshape((train_X.shape[0], train_X.shape[1], 1))

train_y = data[0:data.shape[0],12:data.shape[1]]

neurons = 10
model = Sequential()
model.add(Conv1D(filters=64,input_shape=train_X.shape[1:], 
    activation='relu',kernel_size = 3))
model.add(Flatten())
model.add(Dense(neurons,activation='relu')) # first hidden layer
model.add(Dense(10, activation='softmax'))
sgd = optimizers.SGD(lr=0.05, decay=1e-6, momentum=0.95, nesterov=True)
model.compile(loss='categorical_crossentropy', optimizer=sgd, metrics=['accuracy'])
history = model.fit(train_X, train_y, validation_split=0.2, epochs=1, batch_size=100)

Train on 800 samples, validate on 200 samples
100/800 [==>...........................] - ETA: 2s - loss: 11.4786 - acc: 0.0800
800/800 [==============================] - 0s 547us/sample - loss: 55.3883 - acc: 0.1000 

Answer 2

您需要train_X才能具有第四维。 就像错误消息说的那样。

train_X = train_X.reshape(train_X.shape[0], train_X.shape[1], 1, 1)