0
$\begingroup$

Consider the following problem. The probability of being born in countries [1,2,3,4] is given by [a, b, c, d] respectively. This is a categorical problem.

Now, assume that the height of a person belonging to any country is normally distributed.

The task is to sample from the total distribution of countries and heights,

$ p(c,h) = p(c)p(h|c) $

One can use a mixed distribution as a tool to model these scenarios, which are a special type of DGM.

Can someone please tell me where ML comes into play here, and how this is considered an ML technique?

Improve this question
$\endgroup$

1 Answer 1

Reset to default
0
$\begingroup$

Generative models like latent variable models (e.g. VAE) use directed graphical models and these sort of factorizations as a foundation for learning. In VAEs, Neural nets are used to estimate posteriors/priors to generate samples.

This sort of explicit factorization is helpful in other generative models as well like autoregressive models which are basically operationalizing the chain rule of probability or bayesian networks which may be more explicit in modeling joint distributions of interest.

I would agree the use of ML often overlaps somewhat confusingly/incorrectly in these contexts, but these factorizations often help in formalizing ML problems and can play a more direct role in (deep) generative modeling.

Improve this answer
$\endgroup$
2
  • $\begingroup$ Thanks sma! I am wondering specifically about the problem I gave, say it's a generative model just for generating data, but nothing else. No Neural networks, no deep learning. I am struggling to see where the 'learning' is specially if one were to make a generative model based on the total probability I provide above. $\endgroup$
    – user1029384756
    Jan 20, 2022 at 17:26
  • $\begingroup$ I would rather say that models, both generative and discriminative ones, can be modelled as Bayesian networks, rather than saying "Generative models like latent variable models (e.g. VAE) use directed graphical models". Also, what do you mean by "autoregressive models" specifically? I don't think you mean AR(0), for instance. So, because it's not clear whic autoregressive models you're referring to, then this sentence "which are basically operationalizing the chain rule of probability or bayesian networks which may be more explicit in modeling joint distributions of interest" is also unclear. $\endgroup$
    – nbro
    Jan 22, 2022 at 8:21

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .