Fast on-line learning of point distribution models
Abdullah A. Al-Shaher, Edwin R. Hancock · 2003
We present a fast procedure for training point distribution models (PDM) using the EM algorithm. Rather than estimating the class means and covariance matrices needed to construct the PDM, the method iteratively refines the eigenvectors of the covariance matrix using a gradient ascent technique. We evaluate the method on the problem of learning class-structure of Arabic characters.