Superlinear convergence of a general algorithm for the generalized Foley-Sammon discriminant analysis (Q364740)

The paper deals with the resulting trace ratio optimization problem associated to the linear discriminant analysis approach for feature extraction and dimension reduction. The main task of the paper is to discuss the convergence behavior of an iterative method proposed in [\textit{H. Wang} et al., Proc. International Conf. on Computer Vision and Pattern Recognition, 1--8 (2007)], when the generic assumption is relaxed. In particular, the authors prove that the iteration converges superlinearly and they claim to be able to characterize all possible limit points of this iterative method. An illustrative numerical example is presented.