I tried to customize my loss function in my auto encoder, the loss function must take into account the result of another dimension reduction (LLE) and the data I pass to the function must be updated to each calculates the loss function, the variables that must change do not change. That's my code, I'm waiting for your answers, thank you.
loss function:
def increment():
global i
i = i+1
return i
def call_loss_lle():
global i #i does not increment
def loss_lle(y_true,y_pred):
global i
global increment
i = increment()
X = x_train[i-1:i,]
lamda = 0.3
z = encoder.predict(X)
encoded = encoder.predict(X)
z = z.reshape((28,3))
y,W = LLE_(encoded.reshape((28,3)),10)
produit = np.dot(W,z)
diff = z - produit
loss_lle = lamda * np.linalg.norm(diff)
cross = K.binary_crossentropy(y_true,y_pred)
return cross + loss_lle
return loss_lle
autoencoder:
from keras.layers import Input, Dense
from keras.models import Model
# this is the size of our encoded representations
encoding_dim = 84
# this is our input placeholder
input_img = Input(shape=(784,))
# "encoded" is the encoded representation of the input
encoded = Dense(encoding_dim, activation='relu')(input_img)
# "decoded" is the lossy reconstruction of the input
decoded = Dense(784, activation='sigmoid')(encoded)
# this model maps an input to its reconstruction
autoencoder = Model(input_img, decoded)
# this model maps an input to its encoded representation
encoder = Model(input_img, encoded)
# create a placeholder for an encoded (32-dimensional) input
encoded_input = Input(shape=(encoding_dim,))
# retrieve the last layer of the autoencoder model
decoder_layer = autoencoder.layers[-1]
# create the decoder model
decoder = Model(encoded_input, decoder_layer(encoded_input))
autoencoder.updates()
autoencoder.compile(optimizer='adadelta', loss=call_loss_lle())
