Researches On The Key Technology Of ITO Algorithm

ITO algorithm, a new class of meta-heuristic evolutionary algorithm based on the description of ITO stochastic process of the Brownian motion, has as its key issue the design of drift operator and wave operator. The mapping of ITO stochastic process to the optimization algorithm is able to drive each particle to make search optimization in the local space, this is related to the wave operator design, at the same time, lead particle swarm to move forward in the direction towards the optimal solution, this is concerned with the design of ITO process macro trend term drift operator.ITO algorithm exhibits competitive edge in solving numerical optimization problems. This paper examines its capacity to approach combinatorial optimization problems, taking TSP problems as an example, and then puts forward the neighborhood optimal correlation metrics. After that, it goes on to discuss the design of ITO algorithm drift operator and wave operator, the method of particle radius calculation and the selection of attractive points, and so on. The author also analyzes the algorithm performance values in the case of different parameters and the population diversity. Experimental results show ITO algorithm is well equipped in dealing with symmetric TSP problems and, is especially the case when combined with local search algorithms.The paper presents a theoretical analysis of its convergence and the average expected time to obtain the optimal solution. It demonstrates the strong convergence of ITO through a comprehensive analysis of individual particles and the particle system dynamics, armed with the martingale theory of stochastic processes and Markov processes. It has been found that the expected upper time bound for ITO to solve a class of combinatorial optimization problems can be obtained when it is supported by graph models, and theories of evolutionary algorithm and genetic algorithm analysis, considering the changes of algorithm function in different parameter setting, may well be ported to ITO algorithm, resulting possibly in an enriched version of its convergence theory.An improvement to the description of the algorithm model has been made, relying on the optimized basic mathematical models and processes of ITO algorithm and informed by the ideas of CMA-ES Evolution Strategies, and a new discrete ITO differential formula has been worked out, namely, using ITO's discrete stochastic differential equations to describe its iterative process. With the help of this model, a more efficient ITO algorithm has been designed, and numerical experiments show that the improved algorithm enhances the diversity of particle systems, making it more suitable for solving complex function optimization problems.After an analysis of what might occur when traditional optimization methods are applied to stochastic optimization, the meta-model integration ITO algorithm has been proposed for the function optimization of noise pollution problems. Within this algorithm, the noise deduction can be achieved through the smoothing of the function landscape based on population. In other words, the anticipated outcome does not necessarily mean to run the model several times; it can be attained with the help of the population characteristics of the algorithm. Test results show that for small noise terms, CMA-ES and ITO-MR algorithms function more or less the same, but when moderate and loud noises get involved, the ITO-MR algorithm stands out, with better noise deduction performance, higher accuracy, and fast convergence.