Abstract: A sample and class incremental learning algorithm based on hyper ellipsoidal is proposed. For every class, the smallest hyper ellipsoidal that surrounds most samples of the class is ...