Abstract
The distribution shift due to magnetization differences in magnetometers
causes machine learning-based magnetic localization models to be
ineffective for the end user. We present a novel post-deployment
calibration mechanism to mitigate this issue.
, 
