Local-Global Representation Alignment for Robust Time Series Classification

The proposed LogoRA framework employs a two-branch encoder to extract both local and global representations from time series data, and utilizes various alignment strategies to learn domain-invariant features for robust time series classification across different domains.