Add a tutorial NB for BART

[twiecki]

Currently we only have the doc string, but a good example NB on what it is, how to use it etc would go a long way.

[twiecki]

This could be a good start https://github.com/adamlauretig/pymc_bart/blob/master/pymc_bart.ipynb. @adamlauretig any interest in contributing this to https://github.com/pymc-devs/pymc-examples/?

[adamlauretig]

@twiecki Sure!

I have questions about the diagnostics, since the n_eff is so small. I'm not sure why, and I'm wary of making this official until I can explain it/how to fix it. Once that's resolved though, I'd be happy to contribute.

[twiecki]

I think that's fine to move ahead with. Most important are better explanations/intuitions and better plots of the results.

[ian-whitestone]

Would be great to have an example for binary classification as well if possible. There's an un-verified example here: https://discourse.pymc.io/t/binary-classification-example-using-the-recently-added-bart-model/6420.

If we do draft something up and release, it would be worth mentioning that we don't support out of sample predictions yet (if that is still true).

[aloctavodia]

I can work on this, but probably not sooner than one or two weeks.

[jlevy44]

Are there any updates here? Looking forward to constructing some hierarchical models with BART, but looking through some of the previous examples, would like to know if effective samples can be increased or out of sample accuracy comparable or exceeding RF can be achieved? Thanks!

[chiral-carbon]

I had been planning to work on creating a notebook for BART and have started a bit of work on it. Is someone else also working on this?

[twiecki]

@chiral-carbon Not that I know of, go ahead, looking forward to seeing what you cook up :).

[chiral-carbon]

Thanks @twiecki, I hope I can do a good job! :eyes: