A New Type Of Intelligent Optimization Method - Benchmarking Learning Algorithm

According to the benchmarking theory in the field of enterprise management, the Benchmark Learning Algorithm(BLA), as a new type of intelligent optimization method, was proposed in this paper to cope with the ubiquitous optimization problems.In BAL, a competitive learning mechanism based on dynamic niches was set up according to the core values of benchmarking. In which the active learning-based search strategy took the place of the traditional passive adaptive search strategy. Consequently, some defects of the existing intelligent optimization methods(EIOMs), for example, the running direction of the search process was indecisive, the EIOMs could not maintain the population diversity and could not be adapt to the dynamic optimization problems, were all got overcome. As a result of the imitation and learning to the benchmark, indivduals in the population are able to approach to the target regions in the solution space and seek out the optimal solutions quickly, in respect that the search behavior of these individuals are no longer apt to be completely passive,self-adaptive and random, but active and direction-oriented. What is more, the formidable problem of maintaining the diversity of population was completely overcame through the self-organizing learning process of the niche system and its friendly interaction with the environment, thus, the exploration and exploitation of the BLA will maintain self-adaptively. And then, BAL is able to accurately detect the slight changes of the environment and track the trajectory of the extreme points in the search space, and thus, BLA is naturally adaptable for the dynamic environment.In this paper, the great impact of the learning strategies and control parameters to BLA were analysed, and on the basis of which the best scheme consist of the learning strategies and control parameters was set. And this is the solid theoretical foundation for the practical application of BLA.Additionally, the main differences between BAL and the EIOMs were analysed and revealed. the comparative experiments in four kinds of optimization problems showed that BAL is not only able to deal with the static and dynamic optimization problems, but is capable of coping with the continuous and discrete optimization problem as well. BLA is robust and able to perform friendly interactive learning with the environment, whose static optimization ability and dynamic tracking ability were far superior to the Simple Genetic Algorithm and the Particle Swarm Optimization and so onBLA, which originated from benchmarking theory of business management, is different from the EIOMs, which stemed from the biological activities of nature. BLA is brand new. In other words, BLA is a newborn member of the family comprising the modern intelligent optimization methods.