Learning to Explain: An Information-Theoretic Perspective on Model Interpretation
This addresses the need for interpretable AI models, particularly in domains requiring transparency, but it is incremental as it builds on existing feature selection and mutual information methods.
The paper tackles the problem of model interpretation by introducing instancewise feature selection, which learns to extract the most informative features for each example to explain predictions, and demonstrates its effectiveness on synthetic and real datasets through quantitative metrics and human evaluation.
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 the input is the model to be explained. We develop an efficient variational approximation to the mutual information, and show the effectiveness of our method on a variety of synthetic and real data sets using both quantitative metrics and human evaluation.