Silun Zhang: Decentralization of General Contraction Mappings and Privacy-protected Network Coordination

Abstract

Emerging applications in IoT and edge computing/learning have sparked massive renewed interest in developing distributed versions of existing centralized algorithms often used for optimization and machine learning purposes. While existing work in the literature exhibit similarities, for the tasks of both algorithm design and theoretical analysis, there is still no unified method for decentralizing algorithms. The first part of this talk will present a general framework for distributing the execution of centralized contraction algorithms over networks in which the required information or data is partitioned between connected parties. It shows that the distributed iterative algorithm, which results from the proposed framework, can retain the convergence rate of the original centralized algorithm.

In addition, privacy is a key bottleneck to hamper the application of IoT systems and prevent more parties involving in distributed computation. Due to the information exchanges between nodes, all the advantages of networked systems are essentially at the cost of sacrificing individual privacy in the network. To address this issue, we proposed a privacy-preserving consensus approach based on a secret sharing scheme, whereby all nodes in a network can achieve an agreement on their states without exposing the states to other nodes. Unlike existing works, the proposed privacy-preserving algorithm is resilient to node failures. When a node fails, the method can rebuild the lost node via the information kept in neighbors, even though no neighbor is allowed to know the exact state of the failing node.