English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Covariate distribution shifts and adversarial perturbations presentrobustness challenges to the conventional statistical learning framework: mildshifts in the test covariate distribution can significantly affect theperformance of the statistical model learned based on the trainingdistribution. The model performance typically deteriorates when extrapolationhappens: namely, covariates shift to a region where the training distributionis scarce, and naturally, the learned model has little information. Forrobustness and regularization considerations, adversarial perturbationtechniques are proposed as a remedy; however, careful study needs to be carriedout about what extrapolation region adversarial covariate shift will focus on,given a learned model. This paper precisely characterizes the extrapolationregion, examining both regression and classification in an infinite-dimensionalsetting. We study the implications of adversarial covariate shifts tosubsequent learning of the equilibrium -- the Bayes optimal model -- in asequential game framework. We exploit the dynamics of the adversarial learninggame and reveal the curious effects of the covariate shift to equilibriumlearning and experimental design. In particular, we establish two directionalconvergence results that exhibit distinctive phenomena: (1) a blessing inregression, the adversarial covariate shifts in an exponential rate to anoptimal experimental design for rapid subsequent learning, (2) a curse inclassification, the adversarial covariate shifts in a subquadratic rate fast tothe hardest experimental design trapping subsequent learning.

Blessings and Curses of Covariate Shifts: Adversarial Learning Dynamics, Directional Convergence, and Equilibria