Improvement Research And Its Application Of Transiently Chaotic Neural Network Model

Combinatorial optimization problem is a NP-hard problem, many problems canbe solved by being transformed into a combinatorial optimization problem inscientific research and project technology. In order to find a rational solution to thisproblem, many excellent algorithms were proposed by many scholars and researchers.However, these algorithms can’t gain the ideal results because of their inherentlimitations and defects, Therefore, the improved algorithms based on optimizationproblem become the hot problems in current research.This article is based on the deep study of the transient chaotic neural network(TCNN),we propose its improved algorithm and applied the algorithm to solve theclassical combinatorial optimization problems,channel assignment problem and thetraveling salesman problem(TSP).based on the thoroughly understanding of theinternal mechanism of chaotic neural network(CNN), dynamical behavior of thetransiently chaotic neural network is discussed.The improvements of two aspects ofthe transient chaotic neural network are proposed,so the modified transient chaoticneural network(MTCNN) is formed.One thing is the improvement of incentivefunction of the chaotic neuron, time varying gain is introduced to make sure that thenetwork has a rich dynamics and algorithm has a optimal solution rate,another thing isthe subsection of the annealing function, so the chaos search stage can keep longer inthe search process of TCNN, which can make full use of chaotic characteristics;while the optimal solution can be quickly converged in the stable convergence stage.After the improvement, improvement is proved to be effective by analyzing thechaotic dynamics of the algorithm.At the same time, the improved algorithm isapplied to solve the channel assignment problem and the city traveling salesmanproblem.The energy function of the combinatorial optimization problem areconstructed. Through the simulation, ideal results can be got, and the results arecompared with the simulation result of Hopfield neural network and chaotic neuralnetworks algorithm,the algorithm has a better robustness.the improvement isefficiency for the algorithm. In the simulation, influence of the different parameters on the network arediscussed, and some simulation data and conclusions are obtained, which would giveparameter selection a reference in the future research.the development of the TCNN isenriched.