改进的混沌Hopfield神经网络盲检测算法

An improved blind detection algorithm of chaos Hopfield neural network

以提高Hopfield神经网络盲检测算法激活函数的灵活性为目标,提出一种在原点附近非线性逼近能力更优的激活函数。针对算法存在陷入局部最优的情况,利用混沌映射优良的遍历性和类随机性,在算法起始点利用混沌产生初始序列,在当前全局最优值不变时进行小幅度混沌扰动,以减少算法的误码性能。仿真结果表明,基于激活函数和混沌映射相结合的改进算法,能够提高神经元输入值敏感区域抗干扰能力,加快收敛速度,提高盲检测性能。

In order to improve the flexibility of the activation function of the blind detection algorithm in Hopfield neural network, an activation function with better nonlinear approximation ability near the origin was proposed. For the case where the algorithm trapped in local optima, utilizing the good ergodicity and randomness of chaos mapping, chaos was used to generate the initial sequence at the starting point of the algorithm, and small-amplitude chaotic perturbation was performed when the current global optimum value was constant, so as to reduce the error performance of the algorithm. The simulation results show that the proposed algorithm reduces the sensitivity of neurons to input values, has strong anti-interference ability and fast convergence speed, and improves the blind detection performance.