Double Backpropagation with Applications to Robustness and Saliency Map Interpretability
Christian Etmann · State and University Library Bremen · 2020
This thesis is concerned with works in connection to double backpropagation, which is a phenomenon that arises when first-order optimization methods are applied to a neural network's loss function, if this contains derivatives. Its connection to robustness and saliency map interpretability is explained.