Incremental learning for pattern classification

In analogy to growth of human beings, machine learning should make additional learning if some new knowledge is obtained. This situation occurs very often in practical problems in which the environment changes over time, e.g., in financial investment problems. The well known back propagation method in artificial neural networks is not effective for such incremental learning. On the other hand, the mathematical programming approach and the hypersphere methods including RBF(Radial Base Function), KNN(k Nearest Neighborhood) and the potential method are promising for this purpose. In this paper, the ability of incremental learning is comparatively discussed among the above methods. Examples in stock portfolio problems show the effectiveness of incremental learning.