一种基于最小二乘法的FPN与规则权值优化新方法

Novel weights optimization method based on least square method for FPN AND-Rule

如何确定模糊产生式规则的各项参数对模糊Petri网的建立具有非常重要的意义,也是目前研究的难点和热点。针对模糊产生式规则中的与规则,提出了一种基于最小二乘法求解最优权系数的训练方法,把权值优化问题演变成通过最小二乘算法求解带约束条件的线性超定方程组的问题。仿真结果表明,该算法的训练精度对样本数的依赖较小,训练模型的预测精度较高。与同类模糊Petri网权值优化算法相比,使用该算法得到的最优权系数计算的输出库所对应命题的可信度能够更准确的逼近真实值。

It is of great significance to determine the parameters of the fuzzy production rules for the fuzzy Petri net modeling,which is also the difficult and hot topic of the current research.Aiming at the AND-rule in fuzzy production rules,a weights optimization algorithm based on the least square method is proposed by converting the optimization problem into the solution of restricted over-determined linear equations.Simulation results show that the training accuracy of this algorithm is less dependent on the sample size,and the prediction accuracy is much higher Compared with other FPN weights Optimization algorithms,the output-place value calculated according to this method is more close to the true value.