... to their parallelism and adaptivity to input data, such learning algorithms and their implementations in neuralnetworks are potentially useful in feature detection and data compression tasks.In ... each of the10 character classes, about 1700 handwritten samples were collected, and the samplemeans and covariance matrices were computed by standard estimation methods. Thecovariance matrices ... solve the eigenvectors by standard numericalmethods. In an on-line data compression application like image or speech coding,the data samplesx(t)arrive at high speed, and it may not be possible...