Magnetic core loss prediction often requires transferring models from one material to another, but new materials may have only sparse measurements. The preprint argues that standard global feature alignment is too blunt because source and target data may not cover the same excitation conditions.
The authors propose matching operating conditions and using partial optimal transport, a technique that aligns only part of the distributions. This is intended to account for differences in excitation coverage and avoid forcing irrelevant source data to match target behavior.
The abstract is brief, so implementation details and experimental results are not available from the announcement. The contribution is conceptual: better transfer by focusing on comparable operating regions rather than global alignment.