On the Power of Source Screening for Learning Shared Feature Extractors

arXiv:2602.16125v1 Announce Type: new Abstract: Learning with shared representation is widely recognized as an effective way to separate commonalities from heterogeneity across various heterogeneous sources. …