Research On Cooperative Bacterial Foraging Algorithm For High-Dimensional And Multi-Objective Optimization Problems

Since the1980s, the industrial production process has become a complex production process with the characteristics of large, continuous and comprehensive. Many optimization problems which exist in actual project can be abstracted as ultra high-dimensional or multi-objective optimization problems, seek a kind of algorithm with intelligent characteristics that can be applied to solve large-scale parallel problems has become a main research goal in many subjects. This article is based on a new intelligent algorithms(BFO) to study high-dimensional and multi-objective optimization problems.Aiming at dimension disaster and variable interdependency problem of ultra high-dimensional optimization, this paper proposes a cooperative bacterial foraging algorithm to solve high-dimensional problem. First based on statistics learning method for grouping optimization variable decompose dimensional, according to its correlation threshold, then propose a brief-space-based bacterial foraging optimization(BSBFO), prove the convergence of the algorithm, and use the algorithm optimize the subproblems within each group, finally proposes the cooperative bacterial foraging algorithm(BSBFO_SL) based on cooperation collaborative framework and the strategy of generating optimal sharing vector. Simulated experiments were conducted on10classical benchmarks and7of CEC08benchmarks with500-dimensional and1000-dimensional. The results demonstrate this algorithm effectively improves the ability of scalable when solving high-dimensional problems.Considering the problems of multiple objectives restrain each other and hard to set evaluation function in multi-objective optimization problems(MOO),this paper proposes a modified bacterial foraging algorithm for MOO. Then this paper propose a search-region-based bacterial foraging optimization with variable population(SRBFO-VP). In the process of optimization, use the strategy of variable population strategies to increase the diversity of solution, make the solution distribution uniformity and use exclude strategy to prevent excessive population. Execute region search strategy for non-dominated solution which are sorted by the number of solution in their dominated region to increase the number of PS after executed SRBFO-VP. The result of Simulated experiments indicate SRBFO-VP is very promising in dealing with non local Pareto front MOO,but have modest effects for mang local Pareto front MOO.