“xy are two hidden variables, z is an observed variable, and z has truncation, for example, it can only be observed when z>3, z=x*y, currently I have observed 300 values of z, I should assume that I can get the distribution form of xy, but I don’t know the parameters of the distribution, how to use machine learning methods to learn the distribution parameters of x and y.” This is just a simple example, in my actual problem, the data I observed is not only one dimension, there are three dimensions, and the hidden variables also have three dimensions, but the hidden variables are not directly observable, and the observed values are the mathematical results of three variable， I don’t know what method to use, especially when the observed data has truncation I don’t know how to deal with it, I have a basic understanding of pytorch and pyro framework.”
have tried pyro and torch but I didn’t get the right answer, especially the truncation processing is a big problem for me, if similar examples, give me a reference, that would be great