A sequential learning algorithm of neural network and its application in crop variety selection

In this paper a new incremental machine learning algorithm was proposed and was applied successfully in selecting varieties of strawberry. The proposed algorithm has the characteristics of learning new knowledge of samples in serials, which is referred to as serial machine learning algorithm (SMLA). SMLA realizes incremental version of learning by limits the weight adjustment under a valid boundary. The algorithm can make new knowledge be learned effectively without the old knowledge being stored and re-processed. The advantage of SMLA is that it can expand the knowledge of the system, improve its intelligence and gradually enhance its adaptability. Copyright (C) 1998 IFAC.