# Self-Supervised Dynamic Networks for Covariate Shift Robustness Source: [Self-Supervised Dynamic Networks for Covariate Shift Robustness]() Collection index: [llms.txt]() Data updated: 2025-10-25T22:19:06.029Z Authors: Tomer Cohen, Noy Shulman, Hai Morgenstern, Roey Mechrez, Erez Farhan Published: 2020-06-06T19:37:20.000Z ## Abstract As supervised learning still dominates most AI applications, test-time performance is often unexpected. Specifically, a shift of the input covariates, caused by typical nuisances like background-noise, illumination variations or transcription errors, can lead to a significant decrease in prediction accuracy. Recently, it was shown that incorporating self-supervision can significantly improve covariate shift robustness. In this work, we propose Self-Supervised Dynamic Networks (SSDN): an input-dependent mechanism, inspired by dynamic networks, that allows a self-supervised network to predict the weights of the main network, and thus directly handle covariate shifts at test-time. We present the conceptual and empirical advantages of the proposed method on the problem of image classification under different covariate shifts, and show that it significantly outperforms comparable methods. ## Recorded benchmark results A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. ## Original sources and implementation [Original paper]() [PDF]()