Fuzzy Robust PCA with Intra-sample Outlier Process

抄録

To make Principal Component Analysis (PCA) robust for intra-sample noise, Torre and Black proposed a general analogue outlier process that provides a connection to robust M-estimation. This paper proposes a fuzzy membership approach based on the least squares criterion and Noise Clustering (NC) by Dave for robustifying PCA to intra-sample outliers.