def filterbank_matrices(self, a, mu_x, sigma2, epsilon=1e-9):
t_a, t_mu_x = align(a, mu_x)
temp = t_a - t_mu_x
temp, t_sigma = align(temp, sigma2)
temp = temp / (t_sigma * 2)
F = torch.exp(-torch.pow(temp, 2))
F = F / (F.sum(2, True).expand_as(F) + epsilon)
return F
def mean_channels(F):
assert (F.dim() == 4)
spatial_sum = F.sum(3, keepdim=True).sum(2, keepdim=True)
return spatial_sum / (F.size(2) * F.size(3))
def kldiv2(p, q):
kl = F.sum(F.softmax(p) * (F.log_softmax(p) - F.log(q))) / p.shape[0]
return kl
def kl_divergence(y, t):
entropy = -F.sum(t[t.data.nonzero()] * F.log(t[t.data.nonzero()]))
crossEntropy = -F.sum(t * F.log_softmax(y))
return (crossEntropy - entropy) / y.shape[0]
def ent(p):
h = -F.sum(F.softmax(p) * F.log_softmax(p)) / p.shape[0]
return h
