Improving performance of deep learning models with axiomatic attribution priors and expected gradients
This work addresses the need for more reliable and interpretable deep learning models in applied machine learning, though it is incremental by building on existing attribution priors with a novel method.
The paper tackled the problem of improving deep learning model performance by incorporating axiomatic attribution priors and a new attribution method called expected gradients, which satisfies interpretability axioms like completeness and implementation invariance. The result showed consistent performance gains across image, gene expression, and healthcare datasets, with implementations made freely available.
Recent research has demonstrated that feature attribution methods for deep networks can themselves be incorporated into training; these attribution priors optimize for a model whose attributions have certain desirable properties -- most frequently, that particular features are important or unimportant. These attribution priors are often based on attribution methods that are not guaranteed to satisfy desirable interpretability axioms, such as completeness and implementation invariance. Here, we introduce attribution priors to optimize for higher-level properties of explanations, such as smoothness and sparsity, enabled by a fast new attribution method formulation called expected gradients that satisfies many important interpretability axioms. This improves model performance on many real-world tasks where previous attribution priors fail. Our experiments show that the gains from combining higher-level attribution priors with expected gradients attributions are consistent across image, gene expression, and health care data sets. We believe this work motivates and provides the necessary tools to support the widespread adoption of axiomatic attribution priors in many areas of applied machine learning. The implementations and our results have been made freely available to academic communities.