基于结果反馈的模糊Petri网学习算法

Learning Algorithm of Fuzzy Petri Net Based on Result-Feedback

针对模糊Petri网模型的复杂结构,在不增加虚库所和虚变迁的情况下改进了模糊Petri网分层算法,从而简化模糊Petri网学习和训练方法。为提高收敛速率,本文从一个全新的角度考虑模糊Petri网的学习和训练,提出了基于结果反馈的模糊Petri网学习的新算法(FBFPN)。该算法通过对纯网进行层次式分层及建立变迁点燃的近似连续函数后,调整权值、变迁的阈值、变迁的可信度的同时又调整输入矢量的多重作用来最小化误差函数。仿真结果分析表明,该算法具有良好的学习效率和泛化能力。

With regard to the complex structure of Fuzzy Petri Net, this paper improved the hierarchical algorithm of Fuzzy Petri Net without increasing the virtual place and virtual transition, thereby simplifying the learing and training of Fuzzy Petri Net. To speed convergence, this paper proposed a new algorithm for the learning of Fuzzy Petri Net based on the results feedback, namely FBFPN. Firstly, this algorithm layered the pure net hierarchically and established the approximate continuous function of the transition firing, then adjusted the weight, the threshold and the credibility ,finally adjusted the input vector to minimize the error function. Simulation results showed that this algorithm has stronger generalization ability and higher learning efficiency.