Question for Forward lld (gaussian prior) and entropy estimation in MMILB Module.

[Columbine21]

Multimodal-Infomax/src/modules/encoders.py

Line 152 in 34f92d2

positive = -(mu - y)**2/2./torch.exp(logvar)

Is "positive" vector (above in line 152) for the p(y|x) ~ N(y|µθ1(x), σ2 (x) I)? where is the -(lnσ + C) items in the probability density function for Normal distribution ?

[Columbine21]

Many thx, and looking forwards to your reply.

[Clement25]

Hi, @Columbine21. For the part of BA bound maximization we refer to the implementation of vCLUB, a similar two-stage training approach but for mutual information minimization instead of maximization, https://github.com/Linear95/CLUB/blob/master/MI_DA/MNISTModel_DANN.py (sorry for pasting the wrong link and we have modified that).

[Columbine21]

Thx a lot.

[yangmiemiemie1]

So where is the -(lnσ + C) items in the probability density function for Normal distribution ?Why is it missing?