Abstract
The Alternating Direction Method of Multipliers (ADMM) is a popular and promising distributed framework for solving large-scale machine learning problems. We consider decentralized consensus-based ADMM in which nodes may only communicate with one-hop neighbors. This may cause slow convergence. We investigate the impact of network topology on the performance of an ADMM-based learning of Support Vector Machine using expander, and mean-degree graphs, and additionally some of the common modern network topologies. In particular, we investigate to which degree the expansion property of the network influences the convergence in terms of iterations, training and communication time. We furthermore suggest which topology is preferable. Additionally, we provide an implementation that makes these theoretical advances easily available. The results show that the performance of decentralized ADMM-based learning of SVMs in terms of convergence is improved using graphs with large spectral gaps, higher and homogeneous degrees.
Highlights
In recent years, the exponential growth of digital data opened new challenges for traditional serial methods concerning computation and storage, or in short scalability
We investigate the impact network topology has on the performance of consensus and distributed learning of Support Vector Machines (SVMs) using decentralized Alternating Direction Method of Multipliers (ADMM)
The results from the previous section suggest that the performance of a distributed ADMM-based SVMs algorithm in terms of the number of iterations improves as the degree of graphs becomes larger, i.e., in the group of expander graphs with the fixed number of nodes, the graphs with higher connectivity exhibit accelerated convergence and the complete graphs outperform the expanders
Summary
The exponential growth of digital data opened new challenges for traditional serial methods concerning computation and storage, or in short scalability. Distributed optimization methods with centralized or decentralized computation strategies are virtually inevitable to tackle the challenges arising from large-scale machine learning problems Because of issues such as failure of the central processing unit, communication and synchronization overhead, scalability or privacy the centralized strategies in which the training data is centrally accessible and the distributed agents communicate with a central node may perform poorly (Yang et al, 2019). Decentralized distributed optimization methods in which peer to peer communication is carried out and no central processing unit is involved play an important role in solving large-scale problems Their importance is amplified in scenarios where data are located on distributed nodes, as in edge computing, or where there are concerns preserving privacy (Forero, Cano & Giannakis, 2010)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have