Euler principal component analysis

<p>Principal Component Analysis (PCA) is perhapsthe most prominent learning tool for dimensionality reductionin pattern recognition and computer vision. However,the 2-norm employed by standard PCA is not robust to outliers.In this paper, we propose a kernel PCA method forfast and robust PCA, which we call Euler-PCA (e-PCA).In particular, our algorithm utilizes a robust dissimilaritymeasure based on the Euler representation of complex numbers.We show that Euler-PCA retains PCA’s desirable propertieswhile suppressing outliers. Moreover, we formulateEuler-PCA in an incremental learning framework which allowsfor efficient computation. In our experiments we applyEuler-PCA to three different computer vision applicationsfor which our method performs comparably with other stateof-the-art approaches.</p>