A very fast incremental neural learning for classification using only new incoming datum and hyper-ellipsoidal function

Abstract

This research proposes a very fast 1-pass-throw-away learning algorithm based on a hyper-ellipsoidal function that can be translated and rotated to cover the data set during learning process. The translation and rotation of hyper-ellipsoidal function depends upon the distribution of the data set. In addition, the versatile elliptic basis function (VEBF) neural network with one hidden layer is proposed. The hidden layer of the proposed neural network is adaptively divided into subhidden layers according to the number of classes of the training data set. Each subhidden layer can be scaled by incrementing a new node to learn new samples during training process. The learning time is O(n), where n is the number of data. The network can independently learn any new incoming datum without involving the previously learned data. Therefore, there is no need to store all previous data in order to mix with the new incoming data during the learning process.