An adaptive back-propagation learning method: A preliminary study for incremental neural networks

The authors apply the concept of minimizing weight sensitivity cost and training square-error functions using gradient descent optimization techniques, and they obtain a novel supervised backpropagation learning algorithm on a biased two-layered perceptron. In addition to illustrating the conflict locality of an inserted training instance with respect to previous training data, they point out that this adaptive learning method can get a network with a measurable generalization ability. This work can also be extended to an incremental network in which no training instances are needed to be remembered