I read about a function in tensorflow, which re-parametrizes number of independent distributions to give a probability distribution https://www.tensorflow.org/probability/api_docs/python/tfp/distributions/Independent Unfortunately, there is no sufficient maths explanation of the problem. Can someone help to explain mathematically how its done, or give some references?
Asked
Active
Viewed 71 times
1 Answers
1
As I understand it, it's just taking a few distributions, and assuming independence, you by definition have a new multivariate random variable with the inherited distribution (again, assuming independence).
The rest is just batch-dimension tricks...
The documentation you referenced actually does explain all this, as well as giving examples.
ShlomiF
- 2,686
- 1
- 14
- 19