改进暂态混沌神经网络在信道分配中的应用

Application in Frequency Assignment Based on Improved Transiant Chaotic Neural Network

研究信道分配优化问题,由于传统迭代过程中存在收敛率低,易于陷入局部最优解等缺点。为改善算法收敛速率和信道分配效果,采用改进的暂态混沌神经网络(MTCNN)。在混沌神经网络的动态特性中采用时变增益,在退火过程中采取分段的退火机制,使得混沌搜索阶段保持较长时间的混沌态,利于进行全局搜索,稳定收敛阶段能够迅速收敛于最优解,提高收敛率。仿真结果表明,改进后的算法能很好地解决信道分配问题。和暂态混沌神经网络及仅分段的暂态神经网络相比,最优解率得到很大的提高,网络收敛速度提高了12%以上。最后,给出了模型参数对网络性能影响的一些结论。

Because of its good dynamic characteristic, Chaotic Neural Network （CNN） is applied to solve the Channel Asignment Problem（CAP）. But the convergence rate is low and easily fall into local optimal solution. In or- der to improve the convergence rate and the result of channel allocation, this paper adopted the improved transient chaotic neural network （ MTCNN ）. Time varying gain was applied in the chaotic neural network dynamic character- istics, and two-stage annealing mechanism was adopted in the annealing process of algorithm, so chaotic state can keep longer in the chaotic search stage, in favour of global search, in the stable and convergent stage it can quickly converge to the optimal solution. The convergence rate was improved. The simulation results show that, the improved algorithm can well solve the channel assignment problem. Compared with the transient chaotic neural network and the only two-stage annealing applied situation, the optimal solution rate has been greatly improved, and the network con- vergence speed is accelerated by more than 12%. Finally, some conclusions about the effect of parameters on the net- work model were summed up.