Gradients of Counterfactuals
This addresses the issue of unreliable gradient-based feature importance in saturated deep networks for practitioners in computer vision, bioinformatics, and NLP, though it is incremental as it builds on existing gradient methods.
The paper tackles the problem of feature importance attribution in deep networks where saturation leads to tiny gradients, by proposing interior gradients from counterfactual inputs. The method is applied to GoogleNet, a virtual screening network, and an LSTM language model, showing improved feature importance visualization and attribution properties.
Gradients have been used to quantify feature importance in machine learning models. Unfortunately, in nonlinear deep networks, not only individual neurons but also the whole network can saturate, and as a result an important input feature can have a tiny gradient. We study various networks, and observe that this phenomena is indeed widespread, across many inputs. We propose to examine interior gradients, which are gradients of counterfactual inputs constructed by scaling down the original input. We apply our method to the GoogleNet architecture for object recognition in images, as well as a ligand-based virtual screening network with categorical features and an LSTM based language model for the Penn Treebank dataset. We visualize how interior gradients better capture feature importance. Furthermore, interior gradients are applicable to a wide variety of deep networks, and have the attribution property that the feature importance scores sum to the the prediction score. Best of all, interior gradients can be computed just as easily as gradients. In contrast, previous methods are complex to implement, which hinders practical adoption.