A flexible annealing strategy for chaotic neural network to maximum clique problem

作者：Yang Gang^*; Yi Junyan; Vairappan Catherine; Tang Zheng

来源：International Journal of Innovative Computing Information and Control, 2008, 4(4): 981-993.

• Hopfield neural network
• annealing strategy
• TCNN
• MNN
• maximum clique problem

A flexible annealing strategy for a chaotic neural network is presented in this paper. This annealing strategy can be embedded easily and efficiently in a chaotic neural network. It can make chaotic neural networks get rich chaotic dynamics at the beginning, and then quickly converge to saturate states. Chaotic dynamics and convergence rate can, be tuned flexibly by the annealing strategy. We improve two chaotic neural networks by using this flexible annealing strategy to solve maximum clique problem. Our improved TCNN with the annealing strategy has fewer iteration steps and more efficient ability to get the optimal or near-optimal solution. A maximum neural network with chaotic dynamics is improved and an annealing chaotic maximum neural network with our annealing strategy is proposed, which, can skip local minima successfully and retain good efficiency and rich chaotic dynamics with convergence rate increasing. Simulations on p-random graph, k-random graph and some graphs of the DIMACS clique instances show that chaotic neural networks with the flexible annealing strategy are superior to the other algorithms in light of the solution quality and executive time.

• 出版日期2008-4