Reliability and trust in distributed information systems

Reliability and trust in distributed information systems

Abstract: The distributed consensus problem is of core importance to multi-agent control and coordination. However, it is well known, that consensus algorithms are vulnerable to malicious activity and that the performance guarantees for the nominal case fail in the absence of reliable cooperation. Many works have investigated the possibility of attaining resilient consensus in the face of malicious agents. The first part of this talk presents a new approach to this problem which leads to the conclusion that, under very mild conditions on the link trustworthiness estimate, the deterministic classical bound of 1/2 of the network connectivity can be improved, and significantly more malicious agents can be tolerated.

The second part of this talk presents a semi-decentralized federated learning (FL) approach wherein clients collaborate in relaying their local updates to a central parameter server (PS). FL algorithms iteratively optimize a common objective function to learn a shared model over data samples that are localized over multiple distributed clients. Thus, they suffer from biases and long convergence times when clients are disconnected from the PS. This talk presents a collaborative relaying approach to ensure that the PS regularly receives local updates of poorly connected clients. By strategically optimizing the consensus-based relaying we guarantee convergence to the optimal solution and minimize the convergence time.

Bio: Michal Yemini is an associate research scholar in the Department of Electrical and Computer Engineering at Princeton University. Prior to that, she was a postdoctoral researcher at Stanford University and a visiting postdoctoral researcher at Princeton University. Michal received the BSc in computer engineering from the Technion - Israel Institute of Technology in 2011, and the PhD in electrical engineering in the joint MSc-PhD track from Bar Ilan University, Israel in 2017.