In [1]:

    

%load_ext autoreload
%autoreload 2
%matplotlib inline


    

In [50]:

    

from keras.models import Sequential, Model
from keras.layers import Flatten, Dense, Dropout, Reshape, Permute, Activation, Input, merge
from keras.optimizers import SGD, adam
import numpy as np

from keras.utils import np_utils
from keras.datasets import mnist
from ppap.layers.ppconv import PPConv

(X_train, y_train), (X_test, y_test) = mnist.load_data()
X_train = np.expand_dims(X_train, axis=1)
y_train = np_utils.to_categorical(y_train, 10)
X_test = np.expand_dims(X_test, axis=1)
y_test = np_utils.to_categorical(y_test, 10)


    

In [60]:

    

inputs = Input(shape=(1, 28, 28), name="inputs")
ppconv_1 = PPConv(weight_shape=(3, 3), layer_sizes=[10, 10, 10], nb_filters=2)(inputs)
ppconv_1 =  Activation(activation="relu")(ppconv_1)

ppconv_2 = PPConv(weight_shape=(3, 3), layer_sizes=[10, 10, 10], nb_filters=2)(ppconv_1)
ppconv_2 =  Activation(activation="relu")(ppconv_2)

flat = Flatten()(ppconv_2)
dense_1 = Dense(output_dim=10, activation="softmax")(flat)

model = Model(input=inputs, output=[dense_1])
model.compile(loss="categorical_crossentropy", optimizer="rmsprop", metrics=["accuracy"])

model.summary()
model.fit(x=X_train, y=y_train, nb_epoch=2, verbose=1)


    

    




ola
ola
____________________________________________________________________________________________________
Layer (type)                     Output Shape          Param #     Connected to                     
====================================================================================================
inputs (InputLayer)              (None, 1, 28, 28)     0                                            
____________________________________________________________________________________________________
ppconv_38 (PPConv)               (None, 2, 28, 28)     291         inputs[0][0]                     
____________________________________________________________________________________________________
ppconv_39 (PPConv)               (None, 2, 28, 28)     291         ppconv_38[0][0]                  
____________________________________________________________________________________________________
flatten_18 (Flatten)             (None, 1568)          0           ppconv_39[0][0]                  
____________________________________________________________________________________________________
dense_18 (Dense)                 (None, 10)            15690       flatten_18[0][0]                 
====================================================================================================
Total params: 16272
____________________________________________________________________________________________________
Epoch 1/2
60000/60000 [==============================] - 38s - loss: 14.3739 - acc: 0.1082    
Epoch 2/2
60000/60000 [==============================] - 42s - loss: 13.5661 - acc: 0.1583    






    
Out[60]:





<keras.callbacks.History at 0x7f6d46621fd0>