-
Notifications
You must be signed in to change notification settings - Fork 26
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Pass Cholesky to multivariate gaussian #5
Conversation
Comment by jucor from Saturday Nov 09, 2013 at 08:15 GMT The more I think of it, the more I want it :-) |
Comment by d11 from Monday Nov 11, 2013 at 10:36 GMT Hmm, if we went with the latter, it'd look like the following, right? multivariateGaussianRand( n | resultTensor, mu, { cholesky = myChol }) another option could be multivariateGaussianRand( n | resultTensor, mu, cholesky, true ) where the last argument is a flag indicating that the covariance matrix is already decomposed… On 9 Nov 2013, at 08:15, Julien Cornebise notifications@github.com wrote:
|
Comment by d11 from Monday Nov 11, 2013 at 10:39 GMT On second thoughts, the 'flag' option doesn't work so well if we want to add support for passing the precision matrix too. |
Comment by jucor from Monday Nov 11, 2013 at 10:39 GMT I was thinking a hybrid of both:
where options is a table with
|
Comment by d11 from Monday Nov 11, 2013 at 10:40 GMT I see - yes, that could work quite well I think. On 11 November 2013 10:39, Julien Cornebise notifications@github.comwrote:
|
Comment by jucor from Monday Nov 11, 2013 at 10:41 GMT Yes, exactly: so we can have
where options would have a flag for cholesky and a flag for precision -- thus allowing to pass the inverse of the Cholesky (although precision and cholesky-precision is more useful for logpdf rather than for the sampling, but at least we'd have a unified interface). |
Pass Cholesky to multivariate gaussian
Nice! Thanks! We should put a note in the docs, if you didn't already |
I'm doing it right now ;-) On 18 Nov 2013, at 18:05, Dan Horgan notifications@github.com wrote:
|
And done: http://jucor.github.io/torch-distributions/#toc_13 On 18 Nov 2013, at 18:05, Julien Cornebise julien@deepmind.com wrote:
|
Awesome :)
|
Issue by jucor from Friday Nov 01, 2013 at 18:30 GMT
Originally opened as google-deepmind/torch-randomkit#5