Recurrent Neural Networks for Local Model Prediction
Local models (Walter, I, et al. International Joint Conference on Neural Networks, vol. 1. (1990) 589-594), consists on dividing the data into homogeneous clusters by Vector Quantization (VQ (Gray, R. M., and Neuhoff, D.L. IEEE Trans. Inf. Theory 44(6) (1998) 2325-2383)) to simplify the prediction task on each cluster and mostly inspired from the Self-Organizing Maps algorithm (SOM (Kohonen, T. Self-Organization and associative memory, 3rd edn. (1989))). Since recurrent neural networks have demonstrated in many times a better results and specially for chaotic time series (Bone, R. Recurrent Neural Networks for Time Series Forecasting. (2000)), we propose in this paper a method to use the Recurrent Neural Networks in the local approach.
Time series prediction Recurrent neural network Self organizing map Back propagation through time Local approaches
Aymen Cherif Romuald Bone
国际会议
The Second International Conference on Cognitive Neurodynamics--2009(第二届国际认知神经动力学会议)
杭州
英文
621-628
2009-11-15（万方平台首次上网日期，不代表论文的发表时间）