Abstract: Primal estimated sub-GrAdient solver for SVM (Pegasos) is a simple and effective iterative algorithm for solving the optimization problem of Support Vector Machine．The method alternates between stochastic gradient descent steps and projection steps to find a hyper-plane that can separate two classes of samples with the maximal margin．But it neglects the data distributions which are also vital for an optimal classifier.We developed a novel algorithm,termed as Parallel Primal Estimated sub-GrAdient Solver for Structural SVM (PSPegasos) by embedding the structural information into the SVM and using the parallel computing framework:MapReduce．This algorithm can take full advantage of the computing and storage capacity of the computer cluster,and be applicable to the optimization problem of the massive data．The algorithm was used to two NASA software module datasets CM1and PC1,and the experimental results show that the algorithm can accelerate the convergence speed,improve the classification performance and be an effective solution to the optimization problem of the massive data.

Key words: Structural support vector machine,Parallel,MapReduce