Parallelization Of Tabu Search And Its Application

Tabu search algorithm,as one of many optimization algorithms,has been proposed by Professor Fred Glover since 1986.Because of its unique "memory",it is unique in intelligent algorithms and it has been studied and improved by many scholars,and has been advancing with the times.It involves many disciplines such as electronic science etc.it is widely used to solve vehicle routing problems,scheduling problems,network routing problems,function optimization and other problems,and has achieved ideal results.Tabu search algorithm is an embodiment of artificial intelligence and an extension of local domain search.Through the research status of tabu search at home and abroad,we can see that the biggest disadvantage of tabu search is that the movement of algorithm in solution space is a single "serial movement",unlike other intelligent algorithms,which have inherent parallelism.In this paper,the serial tabu search algorithm is studied deeply,and the basic idea of the algorithm,and the setting and the chosing of key parameters are introduced.On this basis,this paper improves the serial tabu search algorithm,and proposes a parallel tabu search algorithm:First,a parallel strategy is proposed.The parallel strategy is used to add multiple parallel initial solutions to solve the serial tabu search algorithm's initial Solution to the shortcomings of strong dependence.The main idea of it is: the master process is mainly responsible for the initialization of parameters and the decomposition of the original task into several semantically equivalent sub-tasks and sending data to the slave process;The slave process is mainly responsible for receiving the data sent from the master process,completing the local tabu search calculation and transmitting the results back to the master process;Secondly,the correction algorithm is added to effectively eliminate the problem of the existence of a blind spot in the neighborhood solution of the tabu search algorithm;Then,by designing the neighborhood shrinkage factor,the neighborhood space can be adaptively adjusted to enhance the algorithm's search for good solutions.A continuous multi-peak function optimization simulation experiment was carried out on the proposed parallel tabu search algorithm to verify the feasibility and stability of the algorithm.Artificial neural network is a research hot-spot in the field of artificial intelligence.It has successfully solved many practical problems that are difficult to be solved by modern computers in the fields of pattern recognition,intelligent robot,automatic control,prediction and estimation,biology,medicine,economy,etc.,and has shown good intelligent characteristics.In this paper,the parallel tabu search algorithm is used to optimize the connection weights of each layer of the neural network in order to solve the problems of the most widely used and representative BP neural network in which the samples can not be fully trained,easy to fall into local minimum,low generalization ability and strong dependence on the initial solution.MATLAB and Its Parallel Computing Toolbox are used to simulate the problem of speech signal feature recognition and classification and function approximation.The analysis of the experimental results shows that the neural network based on parallel tabu search algorithm has the characteristics of strong global optimization ability,high convergence efficiency,good convergence performance and strong robustness.In addition,the impact of the number of parallel processes on the parallel efficiency is also verified and discussed.Parallel efficiency does not increase blindly with the increase of the number of parallel processes,so when dealing with actual problems,the process should be selected according to the size of the problem.To achieve higher parallel efficiency.