Improvement Of Evolution On Evolutionary Set Theory Based On Public Goods Game Using CUDA

There are many kinds of complex networks in real life or scientific research, all thecomplicated relations can be abstracted as a complex network model which is easily toanalyze and research. With the development of computer technology, there is a huge changewhich can be used to make some corresponding speculation on real world’s evolutionbetween researching and processing on data information.The traditional network structure does not meet the requirements of current networktopology with complex relationship, so this paper based on evolutionary set theory is workingon simulating a real world network evolutionary process and analyzing influence of individualactivities on the evolution of the entire network and looking for the key factors which changethe evolution state through the network degree distribution and individual states.A lot of parallelization technology methods have been used when people not meet theresearch with the traditional ways, with the rapid rise of GPU parallelization and computercluster technology, many researches in some domains can help us to do a better understandingof complex system.A new evolutionary model of network is presented in this paper, which is based on thepublic goods game for simulating a closer model to the real network structure by changing theproportion of collaborators, traitors, average path length and clustering coefficient by differentmultiplier. Individual can changes the strategy for adapting the changing of environment byjoining set, leaving set or imitating strategy of neighbors, and only the strongest one willsurvive in this model.By adjusting the parameters degree distribution of initial stage is close to the Poissondistribution on Small World, then get close to power law distribution of Scale-free network,and in stable condition in the late evolutionary process. Cooperative behavior of theindividuals in the network model for the evolution of the network has a key role is been foundthrough experimental analysis in the evolution found based on the Public Goods Game theory.Cooperative behavior can make a higher clustering coefficient which promotes the networkmodularity aggregation. Individual personal behavior, such as join, exit and imitation wouldaffect the evolution of the network: the join action makes the network together by increasingthe individual activity frequency; the exit do good job to accelerate the steady state process;the imitation changes strategy of the individual. In addition, multiplication factor and thekey parameters of the initial number of connections can be decided to a final state of network evolution, a more attractive set usually has a higher multiplication factor that individualsagainst certain betrayal intrusion behavior under the temptation of higher multiplication factorwhich is crucial to maintaining the cooperation of the entire network evolution; individualshave more frequent communication because there is a higher spread speed in the set withhigher initial number of connections. This paper proposes a new parallel algorithm to simulatethe evolution and development of network based on CUDA which has a distinct acceleratingwith traditional compute way.In conclusion, a new parallel algorithm is proposed in this paper which simulate thedynamic evolution process of complex network, it provide a new research direction for theresearching on this real network.