论文信息 - Stationarity and Stability of Autoregressive Neural Network Processes

Stationarity and Stability of Autoregressive Neural Network Processes

We analyze the asymptotic behavior of autoregressive neural network (AR-NN) processes using techniques from Markov chains and non-linear time series analysis. It is shown that standard AR-NNs without shortcut connections are asymptotically stationary. If linear shortcut connections are allowed, only the shortcut weights determine whether the overall system is stationary, hence standard conditions for linear AR processes can be used.

[1] B. Mandelbrot,et al. Fractional Brownian Motions, Fractional Noises and Applications , 1968 .

[2] David Roxbee Cox,et al. Nonlinear autoregressive processes , 1978, Proceedings of the Royal Society of London. A. Mathematical and Physical Sciences.

[3] H. Tong,et al. On the use of the deterministic Lyapunov function for the ergodicity of stochastic difference equations , 1985, Advances in Applied Probability.

[4] H. Tong. Non-linear time series. A dynamical system approach , 1990 .

[5] Richard A. Davis,et al. Time Series: Theory and Methods , 2013 .

[6] Walter Willinger,et al. On the self-similar nature of Ethernet traffic , 1993, SIGCOMM '93.

[7] Chng Eng Siong,et al. Gradient radial basis function networks for nonlinear and nonstationary time series prediction , 1996, IEEE Trans. Neural Networks.

[8] Sheng Zhaohan,et al. Asymptotic Stationarity of Discrete-time Stochastic Neural Networks. , 1996, Neural networks : the official journal of the International Neural Network Society.

[9] Dirk Husmeier,et al. Predicting Conditional Probability Densities of Stationary Stochastic Time Series , 1997, Neural Networks.