Is R h2o.klime() at all functional yet, or should I hold off and continue to wait?

[Dan]

Hi,
I've been eagerly awaiting H2O's implementation of a version of the LIME algorithm ("k-LIME") since I heard that it was being worked on (previous thread: https://groups.google.com/forum/#!searchin/h2ostream/lime|sort:relevance/h2ostream/Z0ZhTr_lXIY/9zkJFU1dAgAJ).

Recently, I noticed that the newest builds of the R package included a h2o.klime() function. However, after fitting a classifier and attempting to use h2o.klime() I got an error ("java.lang.AssertionError: Unregistered algorithm klime").

My question is: is an error like the one I received to be expected because the h2o.klime() functionality has not yet finished being implemented, or is h2o.klime() finished and working (which would indicate that I received the error due to an error in my own script, which is definitely very possible).

Thanks,
Dan

[mic...@0xdata.com]

Hi Dan,

what you see in the R package is leftover functionality of an experimental feature we were playing with during the development cycle. In the end we decided not to include the feature in the release, unfortunately the API functions didn't get deleted. I am sorry about the confusion!

Michal Kurka

[Dan]

No problem, thanks for letting me know!

[qbitsu...@gmail.com]

it shoud be possible to do this in python but R would be preferable

from __future__ import print_function
import sklearn
import sklearn.datasets
import sklearn.ensemble
import pandas as pd
import numpy as np
import lime
import lime.lime_tabular

R is here https://www.rdocumentation.org/packages/h2o/versions/3.10.5.2/topics/h2o.klime

[Erin LeDell]

that was an experimental feature that was removed.  we recommend using
the lime R or Python package instead.

--
Erin LeDell Ph.D.
Chief Machine Learning Scientist | H2O.ai

[Sathya R]

So,now also H2o doesn't contains KLIME ?

[Michal Kurka]

Hello Sathya,

thank you for your question. K-LIME algorithm is not part of H2O open-source offering (https://github.com/h2oai/h2o-3), it was only included in early experimental/development builds. Currently, there is no plan to introduce K-LIME into H2O OSS. Our focus is directed towards adding interpretable ML algorithms, recent additions include

- RuleFit: https://docs.h2o.ai/h2o/latest-stable/h2o-docs/data-science/rulefit.html

- Generalized Additive Models: https://docs.h2o.ai/h2o/latest-stable/h2o-docs/data-science/gam.html

You can also use our explainability features to get insights into our models: https://docs.h2o.ai/h2o/latest-stable/h2o-docs/explain.html

Thank you,

Michal Kurka

--
You received this message because you are subscribed to a topic in the Google Groups "H2O Open Source Scalable Machine Learning - h2ostream" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/h2ostream/a4sNlVwWLvE/unsubscribe.
To unsubscribe from this group and all its topics, send an email to h2ostream+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/h2ostream/af055c91-d822-4498-941a-5a96291d2b16n%40googlegroups.com.