A Simple Nadaraya-Watson Head can offer Explainable and Calibrated Classification

In this paper, we empirically analyze a simple, non-learnable, and nonparametric Nadaraya-Watson (NW) prediction head that can be used with any neural network architecture. In the NW head, the prediction is a weighted average of labels from a support set. The weights are computed from distances between the query feature and support features. This is in contrast to the dominant approach of using a learnable classification head (e.g., a fully-connected layer) on the features, which can be challenging to interpret and can yield poorly calibrated predictions. Our empirical results on an array of computer vision tasks demonstrate that the NW head can yield better calibration than its parametric counterpart, while having comparable accuracy and with minimal computational overhead. To further increase inference-time efficiency, we propose a simple approach that involves a clustering step run on the training set to create a relatively small distilled support set. In addition to using the weights as a means of interpreting model predictions, we further present an easy-to-compute "support influence function," which quantifies the influence of a support element on the prediction for a given query. As we demonstrate in our experiments, the influence function can allow the user to debug a trained model. We believe that the NW head is a flexible, interpretable, and highly useful building block that can be used in a range of applications.