用FPGA实现的复合神经网络自学习算法电路

Circuit for Self Learning Algorithm of Complex Neural Networks Using FPGA

介绍复合神经网络模型及特性，讨论了基于假逆矩阵的自学习算法，研究用现场可编程门电路（ＦＰＧＡ）实现复合神经网络的自学习过程，电路设计成半并行全数字式．实验结果表明其权值学习结果与计算机软件模拟结果一致。

In this paper,the model of Complex Neural Network (CNN) and its performances are briefly introduced. A pseudo inverse matrix learning algorithm for CNN is discussed. The realizaton of self learning algorithm with Field Programmable Gate Arrays(FPGA) technique is studied. The circuit design is characterized by a completed digitization with a serial parallel combined structure. The experiment results prove that the conclusions are the same as those drawn from software simulation, while its calculation speed is much more outstanding as compared with software simulation.