Distributed support-vector-machine over dynamic balanced directed networks
This work addresses distributed binary classification for networks with privacy constraints, but it is incremental as it builds on existing distributed SVM methods by adding hybrid dynamics.
The paper tackles the problem of training a distributed SVM classifier across a network of agents with limited data sharing, proposing a continuous-time algorithm that handles dynamic network topology changes to eliminate chattering. The result is proven convergence to the SVM classifier over time-varying balanced directed graphs using matrix perturbation theory.
In this paper, we consider the binary classification problem via distributed Support-Vector-Machines (SVM), where the idea is to train a network of agents, with limited share of data, to cooperatively learn the SVM classifier for the global database. Agents only share processed information regarding the classifier parameters and the gradient of the local loss functions instead of their raw data. In contrast to the existing work, we propose a continuous-time algorithm that incorporates network topology changes in discrete jumps. This hybrid nature allows us to remove chattering that arises because of the discretization of the underlying CT process. We show that the proposed algorithm converges to the SVM classifier over time-varying weight balanced directed graphs by using arguments from the matrix perturbation theory.