Real-time novelty detector of chaotic time series based on retina neural network

Abstract

A kind of novelty detection method based on retina neural network is proposed, which could find the novelty in chaotic time series. To demonstrate the capability of the novelty detection method, we designed three novelty detectors,namely the neural network novelty detector (RNNND), back-propagation(BP) novelty detector (BPND) and radial base function(RBF) novelty detector (RBFND), which are based on retina neural network, BP neural network and RBF neural network, respectively. Using Lorenz time series and oil pipeline pressure time series, we tested the performance of the three novelty detectors, including performances of anti-jamming, micro-novelty detection and the computing speed. The results show that the three novelty detectors have good precision and fast computing speed. Finally, the merits and shortcomings of the proposed novelty detection method are analyzed based on retina neural network, BP and RBF neural network, and their applicabilities are given.