SPEAKER: Przemek Biecek; Warsaw University of Technology and University of Warsaw
TITLE: “Black-box openers: How to explain predictions from complex ML models?”
ABSTRACT: Predictive modeling is invaded by elastic, yet complex methods such as neural networks or ensembles (model stacking, boosting or bagging). Such methods are usually described by a large number of parameters or hyper parameters - a price that one needs to pay for elasticity. The very number of parameters makes models hard to understand.
In this talk I will present and compare collection of explainers for complex predictive models, like Break Down (https://arxiv.org/abs/1804.01955), LIME (https://arxiv.org/abs/1602.04938), Ceteris Paribus (https://github.com/pbiecek/ceterisParibus), Shapley Values (https://github.com/slundberg/shap), auditor (https://github.com/mi2-warsaw/auditor/) and DALEX (https://arxiv.org/abs/1806.08915).
Each explainer is a technique for exploration of a black box model. Presented approaches are model-agnostic, what means that they extract useful information from any predictive method despite its internal structure.
It's a part of quickly growing area of research for XAI (Explainable Artificial Intelligence).
STA 290 Seminar List: https://statistics.ucdavis.edu/seminars