混沌神经网络的信息搜索

Information searching in chaotic neural network.

混沌神经网络具有动态联想记忆功能 ,但其动态联想记忆是在网络的混沌运动中实现的 ,记忆状态为连续改变的非周期行为 ,不能稳定在某个已存储的模式上 ,因而混沌神经网络的信息搜索成为一个难点 .本文通过在部分神经元上施加合适的控制量进行钉扎控制 ,改变混沌神经网络的动力学特性 .计算机仿真实验表明 ,在钉扎控制作用期间 ,混沌神经网络的输出能稳定地收敛于网络的存储的模式上 。

The chaotic neural network (CNN) constructed with chaotic neuron shows a dynamical associative memory function. Since its dynamical associative memory is realized in the chaos motion of CNN, the memory states are non-periodic and in continuous changing, the out put of the network can not converge in a stored pattern. So it is difficult to carry out information searching in the CNN. The pinning control method is used for the CNN by adding a suitable control value on parts of neurons. Thus, the dynamic property of the network is changed. The computer simulation demonstrated that the output of the CNN could be converged stably in a stored pattern under the pinning controlling period. The information searching in CNN is achieved.