We describe a gradient-based method to discover local error maximizers of a deep neural network (DNN) used for regression, assuming the availability of an "oracle" capable of providing real-valued supervision (a regression target) for samples. For example, the oracle could be a numerical solver which, operationally, is much slower than the DNN. Given a discovered set of local error maximizers, the DNN is either fine-tuned or retrained in the manner of active learning.
@article{arxiv.2107.13124,
title = {Robust and Active Learning for Deep Neural Network Regression},
author = {Xi Li and George Kesidis and David J. Miller and Maxime Bergeron and Ryan Ferguson and Vladimir Lucic},
journal= {arXiv preprint arXiv:2107.13124},
year = {2021}
}