An information-theoretic perspective on model interpretation

In the ninth seminar of our XAI series, Kristof Schröder, Senior Research Engineer at appliedAI, will discuss how maximizing mutual information between selected features and the response variable can aid in model interpretation, by offering a unique, information-theoretic perspective on AI models.

Abstract: Providing explainability in a model-agnostic way is a challenging task. We have a look at a work by [Che18L], known as L2X, which tackles the problem from an information-theoretic point of view. The key idea is to stack an explainer model and a variational approximation to maximize a relaxation of the mutual information objective, which allows for instance-wise feature selection.

References

[Che18L]

Learning to Explain: An Information-Theoretic Perspective on Model Interpretation, Jianbo Chen, Le Song, Martin J. Wainwright, Michael I. Jordan.

Jun 2018

We introduce instancewise feature selection as a methodology for model interpretation. Our method is based on learning a function to extract a subset of features that are most informative for each given example. This feature selector is trained to maximize the mutual information between selected features and the response variable, where the conditional distribution of the response variable given …

References

In this series →