Incorrect prior_association() ?

[sappersapper]

It seems the method prior_association() can not back propagate gradient to train Ws, or maybe I misunderstood?

anomaly_transformer_pytorch/model.py

Lines 41 to 45 in 6d15200

	def prior_association(self):
	p = torch.from_numpy(
	np.abs(np.indices((self.N, self.N))[0] - np.indices((self.N, self.N))[1])
	)
	gaussian = torch.normal(p.float(), self.sigma[:, 0].abs())

according to paper, is this the right way?
gaussian = 1 / math.sqrt(2 * math.pi) / self.sigma * torch.exp(- 0.5 * (p / self.sigma).pow(2))

[sappersapper]

Does the self.sigma need to be constrained positive? How to constrain it?

[DanyaXu]

Yes, I found the same problem. gaussian = 1 / math.sqrt(2 * math.pi) / self.sigma * torch.exp(- 0.5 * (p / self.sigma).pow(2)).And I emailed the author about it.

I also encountered the same problem, did the author reply to you?