A Non-random Multi-objective Cat Swarm Optimization Algorithm Based On Cat Map And Its Aplication

The problems of multi-objective(MOP),especially in practice and scientific research,are usually challenging and difficult.The research has attracted scholars at home and abroad and gradually become a subject of intense interest.It is much simpler and more effective to apply intelligent algorithm into optimal MOP.This paper proposes a new multi-objective evolutionary algorithm by extending the existing CSO and using the algorithm to address real issues.The model of CSO is built by imitating the natural behavior of cats.Due to its simple concept,high speed of convergence and strong stability,CSO has been applied to many areas such as image processing,pattern recogniti on and networks training,etc.However,in-depth study in theoretical analysis and practical application are still necessary to be carried out for the concept of CSO was put forward late.In order to improve algorithm's efficiency,the paper introduces elite strategy based on the standard CSO,extends CSO to multi-objective optimization area and then uses it to the improved model of Pulse-Coupled Neural Network(PCNN)to realize image segmentation.The main contributions and innovations of this paper can be summarized as follows:1.A new multi-objective optimization,which is A Non-Rando m Multi-Objective Cat Swarm Optimization Algorithm Based on Cat Map(NRC-MOCSO),is put forward.As to shortcomings of CSO,which is easy to fall into local optimum in whole iterative process and has a low convergence rate in the later iterative process,this paper improves CSO,making the cat of population into seeking mode or tracing mode non-randomly.Moreover,the problem that non-unifor m distribution of population on the initialization phase caused instability of CSO algorithm is solved by adding cat map with CSO to initialize population,combining cat map with non-randomly CSO and then extending it to multi-objective optimization algorithm.2.The paper uses proposed NRC-MOCSO to optimize a simple improved model of PCNN(ISPCNN)automatically and achieves multi-objective cat swarm optimization algorithm optimize PCNN parameters for the first time.In simulation,six methods have been involved,including CSO and PSO which use entropy as fitness function,CSO and PSO which use connectivity as fitness function,MOPSO and NRC-MOCSO which use entropy and connectivity as two-objective fitness function.These methods are adopted to optimize parameters of ISPCNN and then the optimized ISPCNN will be used to segment some classical images.The simulation results show two points.Firstly,the selection of fitness functions has a great impact on performance of algorithm;secondly,multi-objective algorithms have advantages over single-objective algorithm w hen solving application problems for multiple targets are able to consider the factors more comprehensively.