Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

You can have Bayesian DNNs.


Not easily. I don't think the literature is very incredible on this - how do you define a prior over all of the parameters of an NN?


There’s a vast literature you can read.


I have read some of the literature - ie. the bayes by backprop method, etc.

It doesn't seem like getting a posterior over the parameter space of a neural network is tractable as of now.


It ain't tractable in general either.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: