Research On The Quotient Space Theory Based Dynamic Problem Solving Method And Its Application

In the field of computational intelligence, granular computing is a new way to simulate human thinking to solve complicated problems. Granular computing involves all the theories, methodologies and techniques of granularity, providing a powerful tool for the solution of complex problems, massive data mining, and fuzzy information processing. Hierarchical quotient space model in artificial intelligence is a problem solving model, which imitates the cognitive process, human thought and behavior. A triplet(X,f,T) is used to describe the quotient space model, which X describes a set of the universal,f(.) describes a attribute function, T is the structure. Different from other granular models, quotient space model use topological structure to describe the relationship inX.In recent years, quotient space theory has developed rapidly in multiple areas. In the process of real production(transportation, petrochemical industry, communications engineering), technical manual, constraint condition and environment are ever changing with time and these kinds of problem solving are called "dynamic problem solving". However, the tradition quotient space model is based on the static data or static topology. Therefore, the tradition quotient space model is limited when the environment changed.In this thesis, the tradition quotient space model, advantage, limitations and application are firstly introduced. Then, inspired by Bayesian method and granular model, dynamic hierarchical quotient space model is proposed. Via evaluating the trustworthiness of nodes, this dynamic model is applied in optimum path search algorithm and Cloud resource scheduling system.The detail research contents are as follows:(1)In this section, the theory of quotient spaces synthesis is expanded into multi-level and multi-dimensional quotient spaces synthesis. Then the model based on the multi-level and multi-dimensional quotient spaces synthesis is proposed. The relationship between the operation of fuzzy equivalence relation, metric space synthesis and multi-dimensional quotient spaces synthesis is discussed.(2)Considered the commercialization and the virtualization characteristics of cloud computing, focusing on the problem of high efficiency and effectiveness resource scheduling, this section proposed for the first time an algorithm of resource scheduling based on Fuzzy quotient spare theory. In the resource scheduling process, each virtual machine attribute is abstracted as an attribute information granulation at first Then the multi-attribute information granulation according to their granular weight, which are defined by the user QoS requirement is studied. Combining with the theory of fuzzy quotient space, fuzzy equivalence partition and distance function are given at last. Based on this, the matching of tasks with resources in cloud environment is implemented. The experimental results show that the algorithm can effectively execute the user tasks and increase resource utilization rate.(3)The dynamic problem solving refers to the kinds of problem-solving methods that its technical manual, constraint condition and environmental factor change with time. There is no effective formal theory to deal with such a complicated problem due to the dynamic problem usually has higher computational complexity. To solve problems under dynamic conditions, we extend the traditional theory of quotient space arming atthe topo logical structure changing with time via using for reference of the trust model in sociology. Inspired by Bayesian cognitive model and quotient space theory, we propose a kind of dynamic hierarchical quotient space model by evaluating the trustworthiness of nodes in complex networks, and then this model is applied in optimal path finding. Theoretical analysis and simulations prove that the proposed model can efficiently meet the requirement of dynamic problem solving, sacrificing fewer time costs, and enhancing paths reliability efficiently.(4)Considering at the trust problem existing in cloud computing environment, a Bayesian method based dynamic quotient space model is proposed to quantify and evaluate the trustworthiness of computing nodes, and its mathematical description and implementation is provided. With the characteristics of dynamic, heterogeneity and deception, resource nodes are inevitably unreliable in the Cloud environment. Therefore, punishment mechanism and pruning-filtering mechanism are also given. Then, a Bayesian subjective trust model based dynamic level scheduling algorithm named BST-DLS is proposed by integration the existing DLS algorithm. Theoretical analyses and simulation experimental results prove that the BST-DLS algorithm can efficiently improve the ratio of successful execution, sacrificing fewer schedule length cost.(5)Via researching on the behavior characteristic of nodes, Cloud failures between nodes considered in fault recovery based reliability model are classified into two categories:unrecoverable failures and recoverable failures. Furthermore, the constraints on the recoverability probability and the numbers of recoveries performed can be imposed freely by resource owners. Then, by integrating the existing DLS algorithm, a dynamic level scheduling algorithm considering fault recovery mechanism named FR-DLS is proposed. Theoretical analyses and simulation experimental results prove that the FR-DLS algorithm can efficiently improve the ratio of successful execution, sacrificing fewer scheduling length and time cost.