def get_output(self, train):
import theano.tensor as T
X = self.get_input(train)
input_dim = X.shape
half_n = self.n // 2
input_sqr = K.sqr(X)
b, ch, r, c = input_dim
extra_channels = T.alloc(0., b, ch + 2 * half_n, r, c)
input_sqr = K.set_subtensor(
extra_channels[:, half_n:half_n + ch, :, :], input_sqr)
scale = self.k
norm_alpha = self.alpha / self.n
for i in range(self.n):
scale += norm_alpha * input_sqr[:, i:i + ch, :, :]
scale = scale**self.beta
return X / scale
def sum_mse(y_true, y_pred):
return K.sqr(y_true - y_pred).sum()
def func(y_true, y_pred):
return K.sum(K.exp(-K.sqr(y_true - y_pred)/sigma))
def loss(x):
x_rec, _, _ = self.mcmc_chain(x, nb_gibbs_steps)
return K.mean(K.sqr(x - x_rec))
def sum_mse(y_true, y_pred):
return K.sqr(y_true - y_pred).sum()
def func(y_true, y_pred):
return -K.mean(K.exp(-K.sqr(y_true - y_pred) / sigma), -1)
def func(y_true, y_pred):
return -K.mean(K.exp(-K.sqr(y_true - y_pred)/sigma), -1)
def loss(x):
x_rec, _, _ = self.mcmc_chain(x, nb_gibbs_steps)
return K.mean(K.sqr(x - x_rec))
