Improving Accuracy-Robustness Trade-Off Via Adaptive Smoothing
Improving neural network accuracy & robustness via adaptive smoothing: mixing standard & robust classifier outputs to achieve high clean accuracy while maintaining strong robustness against adversarial attacks.
This is a Plain English Papers summary of a research paper called Improving the Accuracy-Robustness Trade-Off of Classifiers via Adaptive Smoothing. If you like these kinds of analysis, you should subscribe to the AImodels.fyi newsletter or follow me on Twitter. Overview Proposed a method to significantly improve the trade-off between clean accuracy and adversarial robustness in neural classifiers Mixing output probabilities of a standard (high clean accuracy) and robust classifier, leveraging the robust classifier's confidence difference for correct and incorrect examples Theoretica...