# Explanation of JointDistributionSequential function

**URL:** <https://discuss.ai.google.dev/t/explanation-of-jointdistributionsequential-function/28209>\
**Category:** General Discussion\
**Tags:** decision\_trees, help\_request\
**Created:** [November 10, 2022, 11:10am UTC](https://discuss.ai.google.dev/t/explanation-of-jointdistributionsequential-function/28209 "2022-11-10T11:10:50Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![Julian\_CHAMBRIER](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/julian_chambrier/32/11722_2.png) [@Julian\_CHAMBRIER](https://discuss.ai.google.dev/u/Julian_CHAMBRIER)\
**Post date:** [November 10, 2022, 11:10am UTC](https://discuss.ai.google.dev/t/explanation-of-jointdistributionsequential-function/28209/1 "2022-11-10T11:10:50Z")

</div>

Hello,

After reading the Tensorflow documentation for the JointDistributionSequential function, I still can’t understand how it works in the final inference.

Could someone explain to me, from a mathematical point of view, how the inference to the output distribution is possible.

For example, in the example below,

tfd.JointDistributionNamed(dict(  
min=tfd.Normal(loc=1., scale=3.), #min  
mode=tfd.LogNormal(loc=2., scale=4.), #mode  
max=tfd.PERT(67., 69., 80.), #max  
team\_size=lambda max, mode, min : tfd.PERT(min, mode, max)  
))

how is the inference to the Pert distribution possible ? By what principle is it possible to do this?  
Indeed, we start from 3 different distributions and we infer towards the Pert distribution with these distributions as parameters.

Thanks in advance