Feature Selection Based On Stochastic Distributed Greedy Algorithm

With the advent of big data,variable selection as one of the important research contents of big data analysis has attracted more and more attention.Among them,Greedy Algorithm is widely used in variable selection because of its advantages.However,with the continuous improvement of data magnitude and dimension,the time consumption of Greedy Algorithm is also increasing.In order to solve the problem of large data variable selection,Greedy Algorithm and its improved algorithm under different conditions have been studied.Improved greedy al?gorithm based on submodular function and weak submodularity received more attention.Inspired by this,this paper proposes a variable selection method based on weak submod?ularity,which is faster and suitable for larger and more data cases.It's called Stochastic Distributed Greedy Algorithm.It has a wider scope of application and faster running speed than the existing greedy algorithm,this paper proves the theoretical boundary of it.And then contrasts it with other greedy algorithm under submodularity and weak sub-modularity.Finally,through numerical simulation experiments illustrate the effectiveness and efficiency of Stochastic Distributed Greedy Algorithm.The theoretical proof and simulation results show that the proposed algorithm has good effect and efficiency,and is applicable to a wider range,so it is of great theoretical and practical significance.