CAJ | 학술논문

改进Hopfield神经网络( HNN)的激活函数可以提高网络的抗噪能力,但是其收敛速度会大大降低。为了解决改进激活函数后HNN收敛速度较慢的问题,文中提出一种基于双Sigmoid Hopfield神经网络( DSHNN)的盲检测算法。该算法不仅继承了HNN所有的优点,还极大地提高了算法的收敛速度,缩短了运行时间。仿真实验表明,DSHNN算法比HNN算法的抗干扰性能略强,且其能量函数的收敛速度大大提升。
개진Hopfield신경망락( HNN)적격활함수가이제고망락적항조능력,단시기수렴속도회대대강저。위료해결개진격활함수후HNN수렴속도교만적문제,문중제출일충기우쌍Sigmoid Hopfield신경망락( DSHNN)적맹검측산법。해산법불부계승료HNN소유적우점,환겁대지제고료산법적수렴속도,축단료운행시간。방진실험표명,DSHNN산법비HNN산법적항간우성능략강,차기능량함수적수렴속도대대제승。
Improving the Hopfield Neural Networks' ( HNN) activation function could enhance this network's immunity,but its conver-gence speed was greatly reduced. In order to solve the problem of slow convergence of HNN after activation function improved,applied the Double Sigmoid Hopfield Neural Networks ( DSHNN) for blind detection. This algorithm not only inherited all the advantages of the HNN,but also greatly improved the convergence speed and shortened the running time. Simulation results show that,the DSHNN had stronger anti-interference performance than the HNN slightly and enhanced the convergence rate of the energy function greatly.