基于免疫应答的RBF网络在动态多用户检测中的应用

径向基函数神经网络是一种具有局部逼近能力的前向神经网络,可用作码分多址中的多用户检测器,有较好的检测性能;但是目前的RBF网络多用户检测器收敛慢,不能针对实时变化的动态信道作跟踪.针对动态信道中的多用户检测问题,借鉴免疫系统二次应答原理提出了一种基于免疫机制的RBF网络.该方法结合免疫记忆机制和免疫应答机制设计免疫算子,针对动态模型中的稳定部分和突变部分作不同强度的训练;借鉴免疫系统的二次应答机制,较大地降低了算法复杂度,并且收敛快、鲁棒性强,提高了在变化的信道中的多用户检测器的动态追踪能力.

Application of RBF network based on immune response principle in dynamic multiuser detection

The RBF neural network is a feed-forwaxd neural network model with the capacity of local approximation, so it can be used as a multiuser detector in the CDMA. Although it is proved to be able to present a good performance in the detection, its traditional training algorithms have some disadvantages which lie in the slow convergence speed so that it is not able to trace in dynamic environments. Aiming at the problem of dynamic multiuser detector, a novel training algorithm of RBFNN based on immune mechanism inspired by secondary response principles is proposed. This method introduces vaccine library and immune operators on the basis of immune memory mechanism and immune response mechanism, and takes different intensive level of training to the stable part and changing part of the model. This method possessed of characteristics of lower computational complexity, more rapid convergence and greater robustness, so it has better practicability and real-time performance in the CDMA multiuser detector.