Feature Attribution from First Principles
This work addresses the problem of interpretability in machine learning for researchers and practitioners, but it is incremental as it builds on existing axiomatic approaches.
The paper tackles the challenge of evaluating feature attribution methods by proposing a new framework built from first principles, starting with simple indicator functions and extending to complex models, and derives closed-form expressions for deep ReLU networks.
Feature attribution methods are a popular approach to explain the behavior of machine learning models. They assign importance scores to each input feature, quantifying their influence on the model's prediction. However, evaluating these methods empirically remains a significant challenge. To bypass this shortcoming, several prior works have proposed axiomatic frameworks that any feature attribution method should satisfy. In this work, we argue that such axioms are often too restrictive, and propose in response a new feature attribution framework, built from the ground up. Rather than imposing axioms, we start by defining attributions for the simplest possible models, i.e., indicator functions, and use these as building blocks for more complex models. We then show that one recovers several existing attribution methods, depending on the choice of atomic attribution. Subsequently, we derive closed-form expressions for attribution of deep ReLU networks, and take a step toward the optimization of evaluation metrics with respect to feature attributions.