DYNAMICALLY CAPACITY ALLOCATING NEURAL NETWORKS FOR CONTINUOUS LEARNING USING SEQUENTIAL PROCESSING OF DATA

A nonlinear network model with continuous learning capability is described. The dynamically capacity allocating (DCA) network model is able to learn incrementally as more information becomes available and to avoid the spatially unselective forgetting of commonly used learning algorithms for artificial neural networks. These nonlinear network models are compared to other methods on some classification problems and multivariate calibration of spectroscopic data. In the example cases studied the DCA networks are able to achieve performances that are better than or at least equal to the linear and nonlinear methods tested. In addition to the good prediction performance on the test problems, the DCA networks are able to construct the model using only sequential processing of data. This means that the training data can be collected simultaneously while the network model is already in operation.