混沌优化神经网络求解 job－shop调度问题研究

摘要

In view of Job-shop schedule problems , an optimized model of disperse nonlinear feedback neural network is established , and a neural network optimization method with transient chaos is given , which introduces a process of transient chaos to an optimized neural network , making its evolution have flexible dynamics features .With the reduction of self-feedback coefficient , the state trajecto-ry is shown as a typical inverse bifurcation process , converging to a definite nonlinear feedback neural network and providing a globally near-optimal solution .The essence is to improve the global optimization ability by the randomness and ergodicity of chaos searching , so as to avoid sticking into the local minima .Simulation results indicate that the established model and optimized method have better con -vergence and higher optimization ability than the traditional nonlinear neural network optimization method .%针对Job Shop调度问题，建立了离散非线性回馈神经网络优化模型，给出了一种包含暂态混沌过程的神经网络优化方法。通过在优化神经网络中引入一个暂态的混沌过程，使得网络的演化具备更为灵活的动力学特征。网络状态轨迹随着自反馈系数的衰减，表现为一个典型的倍周期逆分叉过程，逐渐趋向于确定性的非线性回馈神经网络，并为其提供了一个接近全局最优点的初值。其实质是利用混沌搜索过程的随机性和状态遍历性，加强神经网络的全局优化能力，避免陷入局部极小点。仿真结果说明本文所建模型和优化方法比传统的非线性神经网络优化方法具有更好的收敛性和更高优化能力。