神经网络算法在自适应卡尔曼滤波器中的应用

Application of Learning Algorithm for Neural Networks in Adaptive Kalman Filter

针对传统神经网络算法速度慢，容易陷入局部极值的缺点，提出将自适应卡尔曼滤波应用于人工神经网络的训练算法中。把前馈网络中的所有权值、阈值作为自适应卡尔曼滤波算法的状态，网络输出为算法的观测。仿真结果表明，该算法比BP算法在收敛速度方面有明显提高。

According to the question that conventional algorithms for feedforward neural network always suffer from slow convergent rate and local convergence, a learning algorithm training a neural network is proposed for adaptive kalman filter in this thesis. It regards all the weight values and threshold values as the sates, and the outputs of the network as the observing values for the adaptive kalman filter （AKF） in the feedforward networks. Simulation results show that the AKF algorithm is evidently superior to BP algorithm in the rapidity of convergence.