基于遗传算法的Hopfield联想记忆吸引域的优化

Optimization for Attraction Domain of Hopfield Associative Memory with Genetic Algorithm

由于一般离散Hopfield神经网络存在很多伪稳定点,使稳定点的吸引域变小,网络很难获得真正的最优解。因此,提出将遗传算法应用到Hopfield联想记忆神经网络中,利用遗传算法对复杂、多峰、非线性极不可微函数实现全局搜索性质,对Hopfield联想记忆吸引域进行优化,使待联想模式跳出伪模式的吸引域,使Hopfield网络在较高噪信比的情况下保持较高的联想成功率。仿真结果证明了该方法的有效性。

Dueing to the existing of fault equilibrium points in the discrete Hopfield neural network, the attraction domain of the equilibrium points is getting smaller, which leads to the difficulty to acquire the best solution for Hopfield neural network. Then proposes a new method to optimize the attraction domain of Hopfield associative memory with Genetic Algorithm. The characteristics of genetic algorithm, global searching of Genetic Algorithm for complex, multi-apices, and nondifferentiable nonlinear function are applied to the Hopficld neural network, which results in the inputting pattern getting away from the attraction domain of fault pattern and the Hopfield neural network associative memory remaining high success rate while the noise to signal ratio is high. The simulation results prove the validity of the algorithm.