Learn++: a classifier independent incremental learning algorithm for supervised neural networks
Robi Polikar, J. Byorick, Stefanie J. Krause, A. Marino, M. Moreton · 2003
A versatile incremental learning algorithm is introduced for supervised neural network type classifiers. The proposed algorithm, called Learn++, exploits the synergistic expressive power of an ensemble of weak classifiers for learning additional information from new data. Learn++ is capable of learning new classes, without forgetting previously acquired knowledge, even when the previously used data is no longer available. Furthermore, Learn++ is independent of the specific type of the classifier, and adds the incremental learning capability to any supervised neural network classifier.