Distribution that takes as parameter another distribution

TensorFlow General Discussion
,
1

Hello,

I would like to know if it is possible, with tensorflow, to make distributions which take in parameter other distributions.

Here is an example:

tfd.PERT(1, tfd.Normal(1, 2), 10)

or tfd.PERT(tfd.PERT(1, 2, 3) , tfd.PERT(20, 23, 25), 100)

I know there is tfd.JointDistributionSequential but I’m not sure if it does the job.

Thank you for you help

3

Hi @Julian_CHAMBRIER ,

To create distributions in TensorFlow Probability that take other distributions as parameters, you can use tfd.JointDistributionSequential or create custom distributions.

If you need more flexibility or want to define more complex hierarchical relationships, you can create custom distributions using TFP’s tfd.Distribution class.

Additionally you can refer to this official Documentations for better understanding and functionality .

Thank You .