ValueError：检查模型目标时出错：预期activation_2具有形状（无，761,1）但是具有形状的数组（1,779,1）

Question

I have the following error statement...

ValueError: Error when checking model target: expected activation_2 to have shape (None, 761, 1) but got array with shape (1, 779, 1)

In errors, I don't know what the number 761 means, my data1 's shape is 779 * 80 , my data3 's shape is 779 * 1 . Thank you for your help!

from __future__ import print_function
from keras.preprocessing.image import ImageDataGenerator
from keras.models import Sequential

from keras.layers import Dense, \
                         Dropout, \
                         Activation, \
                         Flatten

from keras.layers import Convolution1D, \
                         MaxPooling2D, \
                         Convolution2D

from keras.utils import np_utils

import scipy.io as sio
import numpy as np

matfn = 'LIVE_data.mat'

data = sio.loadmat(matfn) 
data0 = data['data']
data1 = np.ones((1, 779, 80))
data1[0, :, :] = data0
data00 = data['label']
data2 = np.ones((1,779,1))
data2[0, :, :] = data00
data000 = data['ref_ind_live']
data3 = np.ones((1, 779, 1))
data3[0, :, :] = data000
batch_size = 64
nb_classes = 30
nb_epoch = 50

X_train = data1
y_train = data3
X_test = data1[0, :]
y_test = data3[0, :]

X_train = X_train.astype('double')
X_test = X_test.astype('double')
X_train /= 255
X_test /= 255

# Convert class vectors to binary class matrices.
Y_train = np_utils.to_categorical(y_train, nb_classes)
Y_test = np_utils.to_categorical(y_test, nb_classes)

model = Sequential()

model.add(Convolution1D(32, \
                        10, \
                        border_mode = 'same', \
                        input_shape = (779, \
                                       80)))
model.add(Activation('relu'))
model.add(Convolution1D(64, \
                        10, \
                        activation='relu'))
model.add(Dropout(0.25))
model.add(Convolution1D(128, \
                        10, \
                        activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(1))
model.add(Activation('softmax'))

# Let's train the model using RMSprop
model.compile(loss = 'categorical_crossentropy', \
              optimizer = 'rmsprop', \
              metrics=['accuracy'])

print("start train")

model.fit(X_train, \
          Y_train, \
          batch_size = batch_size, \
          nb_epoch = nb_epoch, \
          shuffle = True)

print("end")

score = model.evaluate(X_test, \
                       Y_test, \
                       batch_size = 32)

print('Test score:', \
      score[0])
print('Test accuracy:', \
      score[1])

Answer 1

Your model output shape is (779, 1) and same shape is expected in final layer but due to 2 convolution operation that reduced to 761. So by adding border_mode = 'same' in other 2 convolution layer would solve the problem.

You can check in Model summary:

---

Layer (type) Output Shape Param #

conv1d_1 (Conv1D) (None, 779, 32) 25632

---

activation_1 (Activation) (None, 779, 32) 0

---

conv1d_2 (Conv1D) (None, 770, 64) 20544

---

dropout_1 (Dropout) (None, 770, 64) 0

---

conv1d_3 (Conv1D) (None, 761, 128) 82048

---

dropout_2 (Dropout) (None, 761, 128) 0

---

dense_13 (Dense) (None, 761, 1) 129

---

activation_2 (Activation) (None, 761, 1) 0