EcoSta 2017: Start Registration
View Submission - EcoSta2017
A0369
Title: Robust principal expectile component analysis Authors:  Liang-Ching Lin - National Cheng Kung University (Taiwan) [presenting]
Ray-Bing Chen - National Cheng Kung University (Taiwan)
Mong-Na Lo Huang - National Sun Yat-sen University (Taiwan)
Meihui Guo - National Sun Yat-sen University (Taiwan)
Abstract: Principal component analysis (PCA) is widely used in dimensionality reduction for high dimensional data. It finds principal components by sequentially maximizing the component score variance around the mean. However, in many applications, one is interested in capturing the tail variations of the data rather than variation around the center. In order to capture the tail characters, it was previously proposed principle expectile components (PEC) based on an asymmetric $L_2$ norm. We introduce a new method called Huber-type principal expectile component (HPEC) using an asymmetric Huber norm to produce robust PECs. Statistical properties of the HPEC are derived and a derivative free optimization approach, particle swarm optimization (PSO), is used to find HPECs. As illustrations, HPEC is applied to real and simulated data with encouraging results.