XAI in the jungle of competing frameworks for machine learning

https://xkcd.com/927/

We need more adapters not more standards

DALEX and DALEXtra wrap models and create standardized interfaces. Other libraries for XAI, like: ingredients, auditor etc. may work on top of these uniform wrappers.
The modelStudio package creates an interactive dashboard for model exploration. It’s based on DALEX wrappers thus is independent from model internals and can be used to explore R, scikit-learn, keras or h2o models. Here you can play with a demo.

Smart model wrappers

For R models the explain() function creates a wrapper with uniform interface. You can train your model in any framework and later let DALEX/DALEXtra guess how to transform your model into a standardized wrapper. Here is the full example.
For python models one needs to first serialize the model to a pkl file. Then it will be read by a DALEXtra::explain_scikitlearn() function via the reticulate package. Here you will find more examples for python models.
Partial dependency profiles for four models created in four different frameworks. Common interface to models created in different frameworks helps to cross-compare models. Here is the full example.

Summary

Acknowledgments

--

--

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Przemyslaw Biecek

Przemyslaw Biecek

Interested in innovations in predictive modeling. Posts about eXplainable AI, IML, AutoML, AutoEDA and Evidence-Based Machine Learning. Part of r-bloggers.com.