A Study on the Bayesian Recurrent Neural Network for Time Series Prediction

• DOI: 10.5302/j.icros.2004.10.12.1295
• Publication Date: 2010-09-09T02:31:48Z
• Authors: Hong Chan-Young; Junghoon Park; Yoon Tae-Sung; Jin‐Bae Park
• Keywords: Feedforward neural network; Backpropagation; Echo state network

In this paper, the Bayesian recurrent neural network is proposed to predict time series data. A predictor requests proper learning strategy adjust weights, and one needs prepare for non-linear non-stationary evolution of weights. The in paper estimates not single set weights but probability distributions other words, vector as a state space method, its are estimated accordance with particle filtering process. This approach makes it possible obtain more exact estimation aspect architecture, known that feedback structure superior feedforward problem prediction. Therefore, inference, what we call (BRNN), expected show higher performance than normal network. To verify data numerically generated various kinds applied on order be compared. As result, better backpropagation learning, respectively. Consequently, verified reccurent shows prediction result common