update documentation with uninformative example

[RetrospectiveRotations]

still fails because of missing rules for mea/var of Uninformative node

@Nimrais will take a look

also see here ReactiveBayes/ReactiveMP.jl#588

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 80.28%. Comparing base (08d58ca) to head (f7b76d7).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #621   +/-   ##
=======================================
  Coverage   80.28%   80.28%           
=======================================
  Files          26       26           
  Lines        2135     2135           
=======================================
  Hits         1714     1714           
  Misses        421      421           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[bvdmitri]

Is there a corresponding PR for ReactiveMP or BayesBase? Cannot find it

[RetrospectiveRotations]

Is there a corresponding PR for ReactiveMP or BayesBase? Cannot find it

No since the hacky workaround of setting the mean/variance to 0 is not valid. In the linked beta-bernoulli example, it would treat the uninformative entries as 'false' leading to a wrong posterior (i.e., not equal to the prior).

I think it should do the same as 'missing', but without computing predictions (and computation of free energy should still be possible).

[bvdmitri]

Then maybe we drop this new feature? We can always pass missing there, yeah it activates the prediction, but I don't see a big different from what is proposed here (unless I missed it, pun intented)

[RetrospectiveRotations]

Then maybe we drop this new feature? We can always pass missing there, yeah it activates the prediction, but I don't see a big different from what is proposed here (unless I missed it, pun intented)

It is not a new feature since it is already in ReactiveMP. There is also no documentation for it, hence this PR. But maybe I don't understand correctly why it is originally implemented.