5,337
edits
Line 1,056: | Line 1,056: | ||
==GANs and VAEs== | ==GANs and VAEs== | ||
===Pixel-RNN/CNN=== | |||
* Fully-visible belief network | |||
* Each pixel depends on it's adjacent pixels | |||
* Training: | |||
** Decompose likelihood | |||
** <math>P_{\theta}(x) = \prod_{i=1}^{n} P_{\theta}(x_i | x_1, ..., x_{i-1})</math> | |||
;Pros: | |||
* Can explicitly compute P(x) | |||
* Explicit P(x) gives good evaluation metric | |||
;Cons: | |||
* Sequence generation is slow | |||
* Optimizing P(x) is hard. | |||
==Will be on the exam== | ==Will be on the exam== |