Uncertain canonical correlation analysis for multi-view feature extraction from uncertain data streams

Canonical correlation analysis (CCA) is a well-known technique to extract common features from a pair of multivariate data. In uncertain data stream situations, however, it does not extract useful features because of the existence of data uncertainty which is widespread in a variety of applications. This paper describes an uncertain CCA method called UCCA for feature extraction from uncertain multidimensional data streams. By using the information of uncertainty, UCCA can well represent an uncertain linear structure in the projected space. The approach is tested on a variety of real datasets and its effectiveness in terms of multi-view classification based on dimensionality reduction is validated.