Covariate Shift Regularization using Kernel Methods
How covariate shift affects generalization and reliability. Importance weighting and self-attention-based adaptation.
Investigated how covariate shift affects model generalization and predictive reliability under distribution mismatch, and evaluated importance-weighting and self-attention-based approaches for adapting models to changing input distributions. Student project supervised by Prof. Dr. Nicole Mücke (graded 1.0).