RelatIF: Identifying Explanatory Training Examples via Relative Influence
This work addresses the issue of poor explanatory training example selection in machine learning models, particularly for users needing interpretable explanations, though it is incremental as it builds on existing influence function methods.
The paper tackled the problem that influence functions often identify outliers or mislabelled examples as influential, making them poor for explaining model predictions, by introducing RelatIF, a method that separates global and local influence to select more intuitive explanatory training examples, with empirical evaluations showing improved intuitiveness compared to standard influence functions.
In this work, we focus on the use of influence functions to identify relevant training examples that one might hope "explain" the predictions of a machine learning model. One shortcoming of influence functions is that the training examples deemed most "influential" are often outliers or mislabelled, making them poor choices for explanation. In order to address this shortcoming, we separate the role of global versus local influence. We introduce RelatIF, a new class of criteria for choosing relevant training examples by way of an optimization objective that places a constraint on global influence. RelatIF considers the local influence that an explanatory example has on a prediction relative to its global effects on the model. In empirical evaluations, we find that the examples returned by RelatIF are more intuitive when compared to those found using influence functions.