An adaptive wavelet neural network for spatio-temporal system identification

Starting from the basic concept of coupled map lattices, a new family of adaptive wavelet neural networks(AWNN) is introduced for spatio-temporal system identification, by combining an efficient waveletrepresentation with a coupled map lattice model. A new orthogonal projection pursuit (OPP) method,coupled with a particle swarm optimization (PSO) algorithm, is proposed for augmenting the proposednetwork. A novel two-stage hybrid training scheme is developed for constructing a parsimonious networkmodel. In the first stage, by applying the orthogonal projection pursuit algorithm, significant waveletneurons are adaptively and successively recruited into the network, where adjustable parameters ofthe associated wavelet neurons are optimized using a particle swarm optimizer. The resultant networkmodel, obtained in the first stage, may however be redundant. In the second stage, an orthogonal leastsquares algorithm is then applied to refine and improve the initially trained network by removingredundant wavelet neurons from the network. The proposed two-stage hybrid training procedure cangenerally produce a parsimonious network model, where a ranked list of wavelet neurons, according tothe capability of each neuron to represent the total variance in the system output signal is produced. Twospatio-temporal system identification examples are presented to demonstrate the performance of theproposed new modelling framework.

Neural Networks, 23(10), 1286–1299. Elsevier