Distributed Support Vector Machines Over Dynamic Balanced Directed Networks

In this letter, we consider the binary classification problem via distributed Support Vector Machines (SVMs), where idea is to train a network of agents, with limited share data, cooperatively learn SVM classifier for global database. Agents only processed information regarding parameters and gradient local loss functions instead their raw data. contrast existing work, propose continuous-time algorithm that incorporates topology changes in discrete jumps. This hybrid nature allows us remove chattering arises because discretization underlying CT process. We show proposed converges over time-varying weight balanced directed graphs by using arguments from matrix perturbation theory.