Whenever I encounter articles on supervised learning examples are things like regression, classification, object detection, which are obviously ones following frequentist approach.

I've recently studied generative models, which are basically Bayesian approach since it learns whole distribution, not just a parameter, and found some explanation saying that generative models can be trained using supervised learning (i.e, using labels)

Is supervised learning just the concept about the methodology, in which labels are used or not? Are the common examples suggested above just for ease of understanding, but not the concept itself?

  • $\begingroup$ I don't agree that discriminative = frequentist while generative = Bayesian. $\endgroup$
    – Dave
    Feb 9 at 15:20

2 Answers 2


Loosely speaking, supervised learning means that you have labels for the data (all the clustering and regression), while unsupervised learning that you don't (e.g. clustering).

Bayesian inference is just another way of thinking of the models, parameters, and estimating them. So there is Bayesian linear regression, ridge regression, logistic regression, you can use Bayesian estimation for parameters of neural networks, etc. All those are supervised learning algorithms. There are also clustering (unsupervised) Bayesian models. So Bayesian inference has nothing to do with supervised/unsupervised classification of algorithms.


Of course. There are even Bayesian neural networks.

There is nothing about the Bayesian approach that conflicts with the presence of labels in your dataset.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.