【摘要】 本文提出了用一种神经网络模型——平均场理论来建立多向推理专家系统的方法。对于平均场理论的算法,文中给出了一个学习稳定性定理,指出了在一定条件下平均场理论的收敛性。由于平均场理论的互连性,可用它来建立多向推理专家系统。文中给出了用于调整水泥窑煤-生料配比的专家系统的仿真及其结果。更多还原

【Abstract】 This paper discusses an approach of using Mean Field Theory (MFT) to build multidirectional inference expert systems. It gives a theorem of MFT learning stability, mentions that under some conditions MFT learning will converge. Furthermore, because of the interconnectionship of MFT net, it can be used to build the multidirectional inference expert systems. Here we apply MFT to build an expert system used to regulate the coal to raw material ratio in the cement kiln production system. Also, the simulation result is given.更多还原