Forward KL and Type I error

[lucaalbrizzo]

Hi everyone and to the authors thanks a lot for this great library!

Do you know any method to connect the forward KL divergence to the statistical confidence we have in the reconstruction of the target model (e.g. a formula that connects KL to Type I error)?

It is a more "theoretical question" and I am not sure if this the best place to find an answer so please feel free to close this issue if it is not the case.

Thanks a lot in advance
Cheers
Luca