下三角矩阵神经网络逻辑电路的设计

DESIGN OF LOGIC CIRCUITS BASED ON NEURAL NETWORK WITH LOWER TRIANGULAR MATRIX

由于人工神经网络的卓越优点，为制造超高速、高可靠和可编程的数字集成电路提供了新途径．具有下三角形连接矩阵的Ｈｏｐｆｉｅｌｄ模型在同一输入下仅有唯一的平衡点．本文将讨论基于这种网络模型的组合逻辑电路的逻辑设计方法，以最小化神经元个数为目标的启发式优化算法及权电阻网络参数的计算方法．

A structure of combinational logic circuits based on the ANN model is shown and a logic design method is discussed in this paper. It is proved that this design method can be used to synthesize any combinational functions. A concept of combinational Bolean difference and a heuistic optimization algorithm based on the concept, which searches for the approximately optimal order of input variables to minimize the number of neurons in the network, are presented in this paper. An algorithm for designing ANN weight resistors is also given. Computer simulation and circuit experiments show that the design method and algorithm proposed in this paper are correct and effective.