Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Shouldn't use MAP terminology in Bayes risk context (Concept Check 1) #68

Open
davidrosenberg opened this issue Aug 17, 2018 · 1 comment

Comments

@davidrosenberg
Copy link
Owner

Although mathematically finding a MAP estimate involves finding the argmax of a probability mass function (or density function), and that's exactly what we're doing here, I think that terminology is specific to the Bayesian framework.

https://github.com/davidrosenberg/mlcourse/blob/gh-pages/ConceptChecks/1-Lec-Check_sol.pdfhttps://github.com/davidrosenberg/mlcourse/blob/gh-pages/ConceptChecks/1-Lec-Check_sol.pdf

@davidrosenberg
Copy link
Owner Author

davidrosenberg commented Aug 26, 2018

On further reflection... maybe this is fine: say you have a classic latent variable model z --> x, where x is observed, z is latent. When the distributions of z and x are parameterized by a parameter theta, it makes sense to talk about a MAP estimate of theta (in bayesian context) or ML estimate of theta (frequentist context). But I think it would confusing to also talk about the MAP of z.

However, with parameter theta fixed, we have a full joint distribution over z and x, with x observed and z unobserved --- mathematically that's exactly the Bayesian setting. So if there are no parameter to estimate, then it pretty clear what's meant...

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant