Fit interpretable models. Explain blackbox machine learning.
InterpretML - Alpha Release
Equal contributions: Samuel Jenkins & Harsha Nori & Paul Koch & Rich Caruana
In the beginning machines learned in darkness, and data scientists struggled in the void to explain them.
Let there be light.
InterpretML is an open-source package for training interpretable models and explaining blackbox systems. Interpretability is essential for:- Model debugging - Why did my model make this mistake?- Detecting bias - Does my model discriminate?- Human-AI cooperation - How can I understand and trust the model's decisions?- Regulatory compliance - Does my model satisfy legal requirements?- High-risk applications - Healthcare, finance, judicial, ...
Historically, the most intelligible models were not very accurate, and the most accurate models were not intelligible. Microsoft Research has developed an algorithm called the Explainable Boosting Machine (EBM)* which has both high accuracy and intelligibility. EBM uses modern machine learning techniques like bagging and boosting to breathe new life into traditional GAMs (Generalized Additive Models). This makes them as accurate as random forests and gradient boosted trees, and also enhances their intelligibility and editability.
In addition to EBM, InterpretML also supports methods like LIME, SHAP, linear models, partial dependence, decision trees and rule lists. The package makes it easy to compare and contrast models to find the best one for your needs.
*EBM is a fast implementation of GA2M. Details on the algorithm can be found here.
Python 3.5+ | Linux, Mac OS X, Windowsshpip install numpy scipy pyscaffoldpip install -U interpret
Multiple ways to get in touch:- Email us at firstname.lastname@example.org- Or, feel free to raise a GitHub issue
Reporting Security Issues (we had to include this...)
Security issues and bugs should be reported privately, via email, to the Microsoft SecurityResponse Center (MSRC) at email@example.com. You shouldreceive a response within 24 hours. If for some reason you do not, please follow up viaemail to ensure we received your original message. Further information, including theMSRC PGP key, can be found inthe Security TechCenter.
If a tree fell in your random forest, would anyone notice?