search

arviz-devs / preliz

Exploring and eliciting probability distributions
https://preliz.readthedocs.io
Apache License 2.0
90 stars 9 forks source link

Workflow is strange for creating PyMC priors with maximum entropy #587

Open ricardoV94 opened 1 week ago

ricardoV94 commented 1 week ago

One has to pass plot=False all the time which is odd and verbose. Should plot just be an ipython repr thing (like pandas, you won't see the DataFrame unless that's the last command in the cell)

aloctavodia commented 1 week ago

I have being thinking about having rcparams like in arviz. That will allow to globally disable plotting. @OriolAbril what do you think?

ricardoV94 commented 1 week ago

Why should the library plot by default?

aloctavodia commented 1 week ago

I see both options as equally valid. Which is the obvious one and which is the annoying one depends on your workflow. That's why a global variable could be a more flexible approach.

In a more "traditional" elicitation workflow there is a necessity to inspect and reflect on the results before continuing, the process should not be automatic and the human should be in the loop. Many options are evaluated. Then visualizing the results by default makes sense.

In a more automatized workflow where a function like maxent is just one part of the pipe plotting by default becomes super annoying.