RuiShu/vae-clustering

categorical prior loss

patrickltobing opened this issue · 1 comments

patrickltobing commented

Hi, thank you for the great solution.

Though, regarding the prior loss term,
I think the number of mixtures K should be included in labeled_loss(x, px_logit, ...):,
https://github.com/RuiShu/vae-clustering/blob/master/shared_subgraphs.py#L31,
so it will return xy_loss + np.log(K).

sghalebikesabi commented

He correctly writes xy_loss - np.log(0.1) = xy_loss - np.log(1/K) = xy_loss + np.log(K)

  • 👍1