本文提出了一种联想记忆网络中的最速递减动态演化规则，首先改变能产生最大系统能量递减的神经元状态，从而减小落入多余吸引子的可能性．在理论上，我们分析了样本的吸引域和联想规则的收敛性．计算机实验结果说明了它的优越性． In this paper, we propose a dynamic evolution rule of the steepest descent in associative memory networks by first changing the state of neurons that produce the largest system energy decrement, thereby reducing the possibility of falling into excess attractors. In theory, we analyze the convergence of sample attracting domains and associative rules. Computer experimental results illustrate its superiority.