摘要
In this study, a novel adaptive BP learning algorithm with optimal learning rates and directional error correction is proposed. In this algorithm, the optimal adaptive learning rates are used to adjust the weights in an adaptive and dynamical way. Meantime, the directional errors are fed back to learning process to increase the directional accuracy. For illustration and testing purposes the proposed algorithm is applied to the foreign exchange market trend prediction.
- 出版日期2006
- 单位中国科学院数学与系统科学研究院; 香港城市大学; 湖南大学