Selecting the number of components in PCA via random signflips

David Hong, Yue Sheng, Edgar Dobriban

Principal component analysis (PCA) is a foundational tool in modern data analysis, and a crucial step in PCA is selecting the number of components to keep. However, classical selec…