Incremental Learning of Ultrasonic Weld Inspection Signals

In this paper, we present LEARN++, a new algorithm that allows existing classifiers to learn incrementally from new data without forgetting previously acquired knowledge. LEARN++ is based on generating multiple classifiers, each trained with a different subset of the training data and then combining them to form an ensemble of classifiers using weighted majority voting. The fundamental contribution of the algorithm lies in the manner in which the training data is partitioned. The results demonstrate the feasibility and effectiveness of the approach in learning from new data.