Incremental neural learning using AdaBoost

An incremental learning system updates its hypotheses as new instances arrive without reexamining old instances. This paper describes our research in incremental neural learning. We developed an incremental neural learning (INL) framework that allows a neural network system to incrementally learn new knowledge from only new data without forgetting the existing knowledge. We have applied the INL to a vehicle fault diagnostics problem and our experiments showed very positive results.