-
Notifications
You must be signed in to change notification settings - Fork 246
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
icdf of beta distribution is not implemented #1365
Comments
Hi @ayaka14732, please subscribe this upstream issue jax-ml/jax#2399 Just curious, what is your usage case? If you don't need differentiation, then we can use scipy implementation for it. |
Just for learning, so I can use the scipy implementation. But for the NumPyro library, I think we should wait for it to be added in JAX, so users won't experience unexpected performance degradation. |
I'd like to use the beta icdf in a Any recommendations for what would currently be the best / easiest way to circumvent this problem? Can I somehow implement and provide the gradient information by myself? |
I guess if beta icdf is only used at deterministic sites for prediction, then you can use |
(In case somebody else finds this and looks for a solution, I opted to use the Kumaraswamy distribution instead, as recommended in the jax issue linked to above. It's very similar / in some cases identical to the beta distribution and has the icdf implemented.) |
As suggested in jax-ml/jax#2399, we can use |
This can be added in the same way as #1475 . |
NumPyro version: 0.9.1
The text was updated successfully, but these errors were encountered: