Parallel Branch and Bound Algorithms on Semi-supervised SVMs

Branch and bound for semi-supervised support vector machines as an exact, globally optimal solution is useful for benchmarking different practical S3VM implementation. But, global optimization can be computationally very demanding. Parallel implementation of the algorithm enables us to reduce computational time significantly and to solve larger problems. Focusing on the time consuming problem of BBS3VM, a novel parallel branch and bound semi-supervised support vector machines (PBBS3VM) is proposed. Experimental results on server dataset show that the proposed algorithm outperforms the classical sequential algorithm in terms of accuracy and greatly reduce the running time using the Linux PCs.