基于迟滞噪声混沌神经网络的TDMA广播调度

分组无线网络的时分多址(Time Division Multiple Access,TDMA)广播调度问题是一个经典的NP-hard组合优化问题,可用神经网络求解.混沌动力学、随机游动和迟滞动力学均能够有效地提高神经网络的优化性能.为了提高迟滞动力学在噪声混沌神经网络中的优化能力,又不增加噪声混沌神经网络的参数,将噪声混沌神经网络的噪声幅值作为Sigmoid函数的中心参数,并通过神经元的输入变化来控制噪声幅值形成迟滞环,提出了一种新型的迟滞噪声混沌神经网络.对神经元状态演化行为的研究表明,该网络能够同时演化出混沌倒分岔、随机游动和迟滞等动力学行为.对分组无线网络的TDMA广播调度问题的仿真表明,提出的迟滞噪声混沌神经网络具有更好的优化性能.

TDMA broadcast scheduling based on hysteretic noisy chaotic neural network

Time division multiple access(TDMA) broadcast scheduling problem in packet radio networks is a classical NP-hard combinatorial problem,which can be solved by neural networks.Chaotic dynamics,stochastic wandering,and hysteretic dynamics can improve the optimization performance of neural networks effectively.In order to enhance the optimization ability of hysteretic dynamics in the noisy chaotic neural network,and not to increase any parameters into the noisy chaotic neural network,this paper presented a novel hysteretic noisy chaotic neural network by taking noise amplitudes of the noisy chaotic neural network as center parameters of Sigmoid function and using inputs’ change of neurons to control noise amplitudes to form hysteretic loop.Research on state evolutionary of neurons indicates that the proposed network could evolve dynamics including chaotic reverse bifurcation,stochastic wandering and hysteresis.Simulations in TDMA broadcast scheduling problem in packet radio networks suggest that the proposed hysteretic noisy chaotic neural network can behave better optimization performance.