Poster
A Benchmark for Interpretability Methods in Deep Neural Networks
Sara Hooker · Dumitru Erhan · Pieter-Jan Kindermans · Been Kim
East Exhibition Hall B, C #159
Keywords: [ Data, Challenges, Implementations, and Software ] [ Benchmarks ] [ Visualization or Exposition Techniques for Deep Networks ] [ Deep Learning ]
We propose an empirical measure of the approximate accuracy of feature importance estimates in deep neural networks. Our results across several large-scale image classification datasets show that many popular interpretability methods produce estimates of feature importance that are not better than a random designation of feature importance. Only certain ensemble based approaches---VarGrad and SmoothGrad-Squared---outperform such a random assignment of importance. The manner of ensembling remains critical, we show that some approaches do no better then the underlying method but carry a far higher computational burden.
Live content is unavailable. Log in and register to view live content