X_train = X_train.reshape(X_train.shape[0], 1, shapex, shapey)
X_test = X_test.reshape(X_test.shape[0], 1, shapex, shapey)
X_train = X_train.astype("float32")
X_test = X_test.astype("float32")
X_train /= 255
X_test /= 255
print('X_train shape:', X_train.shape)
print(X_train.shape[0], 'train samples')
print(X_test.shape[0], 'test samples')
# 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 = NN_Model(n_epochs=nb_epoch, n_batch=batch_size, val_Freq=1)
model.add(Convolution2D(nb_filters, 1, nb_conv, nb_conv, border_mode='full'))
model.add(Activation('relu'))
model.add(Convolution2D(nb_filters, nb_filters, nb_conv, nb_conv))
model.add(Activation('relu'))
model.add(MaxPooling2D(poolsize=(nb_pool, nb_pool)))
model.add(Drop_out(0.25))
model.add(Flatten())
# the resulting image after conv and pooling is the original shape
# divided by the pooling with a number of filters for each "pixel"
# (the number of filters is determined by the last Conv2D)
model.add(FC_layer(nb_filters * (shapex / nb_pool) * (shapey / nb_pool), 128))
model.add(Activation('relu'))
model.add(Drop_out(0.5))
X_train = X_train.reshape(X_train.shape[0], 1, shapex, shapey)
X_test = X_test.reshape(X_test.shape[0], 1, shapex, shapey)
X_train = X_train.astype("float32")
X_test = X_test.astype("float32")
X_train /= 255
X_test /= 255
print('X_train shape:', X_train.shape)
print(X_train.shape[0], 'train samples')
print(X_test.shape[0], 'test samples')
# 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 = NN_Model(n_epochs=nb_epoch,n_batch=batch_size,val_Freq=1)
model.add(Convolution2D(nb_filters, 1, nb_conv, nb_conv, border_mode='full'))
model.add(Activation('relu'))
model.add(Convolution2D(nb_filters, nb_filters, nb_conv, nb_conv))
model.add(Activation('relu'))
model.add(MaxPooling2D(poolsize=(nb_pool, nb_pool)))
model.add(Drop_out(0.25))
model.add(Flatten())
# the resulting image after conv and pooling is the original shape
# divided by the pooling with a number of filters for each "pixel"
# (the number of filters is determined by the last Conv2D)
model.add(FC_layer(nb_filters * (shapex / nb_pool) * (shapey / nb_pool), 128))
model.add(Activation('relu'))
model.add(Drop_out(0.5))
train=(train_x,train_y)
'''
n_y = np.max((np.max(train[1]),np.max(valid[1]))) + 1
print 'number of classes: %i'%n_y
print 'number of training data: %i'%len(train[0])
print 'number of validation data: %i'%len(valid[0])
####build model
print 'Initializing model...'
mode='tr'
model = NN_Model(n_epochs=n_epochs,n_batch=n_batch,snapshot=snapshot_Freq,
sample_Freq=sample_Freq,val_Freq=val_Freq,L1_reg=L1_reg,L2_reg=L2_reg)
model.add(Embedding(n_words,dim_word))
model.add(Drop_out(0.25))
model.add(GRU(n_u,n_h))
model.add(Drop_out())
model.add(Pool('mean'))
model.add(Drop_out())
model.add(FC_layer(n_h,n_y))
model.add(Activation('softmax'))
model.compile(optimizer=optimizer,loss=loss)
filepath='save/review3.pkl'
if mode=='tr':
