Research On Heuristic Algorithm Of Bipartite Graph Structure

With the rapid development of intelligent information,more and more researchers have focused on heuristic algorithm,which has good performance in many combinatorial optimization problems.However,with the arrival of big data stage,the multi-modal information data and the heterogeneity data object have become difficult problems in reality optimization,the traditional heuristic algorithm haven't not met the needs of real application.Therefore,it is necessary to remodel the optimization problem in reality,combine the relevant theories of mathematical model to analyze the real application problem.At the same time,it‘s meaningful to do research work in improving the performance and enhancing the efficiency of existing heuristic algorithm,including proposing a new search mechanism or a hybrid search algorithm.In order to develop a variety of possible heuristic algorithms,we need to analyze advantages and disadvantages of the present heuristic strategy.Based on the research of existing heuristic algorithms,this paper proposed some heuristic algorithms on combinatorial optimization problem with bipartite graph structure information.Bipartite graph structure is difficult to match with many constraints,many heterogeneous objects and uncertainty.Therefore,a bipartite graph matching framework is proposed based on variable correlation learning.Alpha matting problem and two-echelon vehicle routing optimization problem are regarded as research object,so three heuristic strategies are put forward.The research work is listed as the follows:Firstly,a new heuristic strategy about increasing the diversity of foreground and background color samples has been proposed for alpha matting,which is a combinatorial optimization problem with bipartite graph structure information.In order to increase the foreground and background color sample diversity and avoid missing the true samples during the matching process between foreground sample and background sample,the proposed algorithm makes full use of the correlation between unknown pixels by increasing the self-organizing learning ability with immune algorithm and solving the best match for foreground and background sample pair with parallel particle swarm optimization.The heuristic strategy with added diversity is effective to improve the performance of the complex image of the based-sample matting method,which has laid the foundation for solving the complex combinatorial optimization problem with bipartite graph structure information.Secondly,another heuristic strategy based on personal preference is introduced to solve combinatorial optimization problem with compatible bipartite graph structure,which is effective to improve the co-optimization performance between different heterogeneous data objects in a way of man-machine cooperation.Computer games,as a tool for human-computer cooperation,simulate the realistic scene of optimization problem and collect the personal preferences in the decision-making process.Two-echelon vehicle routing problem is regarded as an example of optimization problem with compatible bipartite graph structure.The proposed method exploits the personal preference on satellite in two-echelon vehicle routing problem for setting out the appropriate allocation strategy on the first echelon and two echelon routing problems.At the same time,personal preferences sets up the bridge between the multi-population artificial ant colonies enhancing the performance of the heuristic algorithm based on the preference strategy.Thirdly,a heuristic strategy method based on fuzzy evolution matching is proposed to solve the uncertainty of matching optimization problem with complex bipartite graph structure information.Based on the heuristic strategy with diversity strategy and the personal preference strategy,the proposed paper analyzes the internal correlation of subset of the bipartite graph for matching optimization problem.The fuzzy relation over subset of the bipartite graph is defined,fuzzy subsets with different relations is obtained by fuzzy matrix decomposition.Combining with the theory of fuzzy sets,the fuzzy matching procedure is proposed enhancing the matching efficiency between different heterogeneous objects,which lay a solid foundation for improving the performance of multi-heterogeneous object matching problem with bipartite graph structure.