quick intro to attribution methods

view markdown

A variety of attribution methods have recently sprung up to try and understand how a fully trained model, makes a prediction on a single datapoint. Here, we provide an extremely quick overview to some of these methods

LIME (ribeiro et al. 2016)

SHAP (lundberg & lee 2017)

This example is based on this wonderful blog post - see there for more details.

ACD/CD (singh et al. 2019, murdoch et al. 2018)