skip to main content


Title: DAve-QN: A Distributed Averaged Quasi-Newton Method with Local Superlinear Convergence Rate
In this paper, we consider distributed algorithms for solving the empirical risk minimization problem under the master/worker communication model. We develop a distributed asynchronous quasi-Newton algorithm that can achieve superlinear convergence. To our knowledge, this is the first distributed asynchronous algorithm with superlinear convergence guarantees. Our algorithm is communication-efficient in the sense that at every iteration the master node and workers communicate vectors of size 𝑂(𝑝), where 𝑝 is the dimension of the decision variable. The proposed method is based on a distributed asynchronous averaging scheme of decision vectors and gradients in a way to effectively capture the local Hessian information of the objective function. Our convergence theory supports asynchronous computations subject to both bounded delays and unbounded delays with a bounded time-average. Unlike in the majority of asynchronous optimization literature, we do not require choosing smaller stepsize when delays are huge. We provide numerical experiments that match our theoretical results and showcase significant improvement comparing to state-of-the-art distributed algorithms.  more » « less
Award ID(s):
1814888 1723085
NSF-PAR ID:
10256971
Author(s) / Creator(s):
Date Published:
Journal Name:
Proceedings of Machine Learning Research
Volume:
108
ISSN:
2640-3498
Page Range / eLocation ID:
965-1976
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Nonconvex and nonsmooth problems have recently attracted considerable attention in machine learning. However, developing efficient methods for the nonconvex and nonsmooth optimization problems with certain performance guarantee remains a challenge. Proximal coordinate descent (PCD) has been widely used for solving optimization problems, but the knowledge of PCD methods in the nonconvex setting is very limited. On the other hand, the asynchronous proximal coordinate descent (APCD) recently have received much attention in order to solve large-scale problems. However, the accelerated variants of APCD algorithms are rarely studied. In this project, we extend APCD method to the accelerated algorithm (AAPCD) for nonsmooth and nonconvex problems that satisfies the sufficient descent property, by comparing between the function values at proximal update and a linear extrapolated point using a delay-aware momentum value. To the best of our knowledge, we are the first to provide stochastic and deterministic accelerated extension of APCD algorithms for general nonconvex and nonsmooth problems ensuring that for both bounded delays and unbounded delays every limit point is a critical point. By leveraging Kurdyka-Łojasiewicz property, we will show linear and sublinear convergence rates for the deterministic AAPCD with bounded delays. Numerical results demonstrate the practical efficiency of our algorithm in speed. 
    more » « less
  2. This papers studies multi-agent (convex and nonconvex) optimization over static digraphs. We propose a general distributed asynchronous algorithmic framework whereby i) agents can update their local variables as well as communicate with their neighbors at any time, without any form of coordination; and ii) they can perform their local computations using (possibly) delayed, out-of-sync information from their neighbors. Delays need not be known to the agents or obey any specific profile, and can also be time-varying (but bounded). The algorithm builds on a tracking mechanism that is robust against asynchrony (in the above sense), whose goal is to estimate locally the sum of agents’ gradients. When applied to strongly convex functions, we prove that it converges at an R-linear (geometric) rate as long as the step-size is sufficiently small. A sublinear convergence rate is proved, when nonconvex problems and/or diminishing, uncoordinated step-sizes are employed. To the best of our knowledge, this is the first distributed algorithm with provable geometric convergence rate in such a general asynchonous setting. 
    more » « less
  3. In many scenarios, information must be disseminated over intermittently-connected environments when the network infrastructure becomes unavailable, e.g., during disasters where first responders need to send updates about critical tasks. If such updates pertain to a shared data set, dissemination consistency is important. This can be achieved through causal ordering and consensus. Popular consensus algorithms, e.g., Paxos, are most suited for connected environments. While some work has been done on designing consensus algorithms for intermittently-connected environments, such as the One-Third Rule (OTR) algorithm, there is still need to improve their efficiency and timely completion. We propose CoNICE, a framework to ensure consistent dissemination of updates among users in intermittently-connected, infrastructure-less environments. It achieves efficiency by exploiting hierarchical namespaces for faster convergence, and lower communication overhead. CoNICE provides three levels of consistency to users, namely replication, causality and agreement. It uses epidemic propagation to provide adequate replication ratios, and optimizes and extends Vector Clocks to provide causality. To ensure agreement, CoNICE extends OTR to also support long-term network fragmentation and decision invalidation scenarios; we define local and global consensus pertaining to within and across fragments respectively. We integrate CoNICE's consistency preservation with a naming schema that follows a topic hierarchy-based dissemination framework, to improve functionality and performance. Using the Heard-Of model formalism, we prove CoNICE's consensus to be correct. Our technique extends previously established proof methods for consensus in asynchronous environments. Performing city-scale simulation, we demonstrate CoNICE's scalability in achieving consistency in convergence time, utilization of network resources, and reduced energy consumption. 
    more » « less
  4. Stochastic Gradient Descent (SGD) is a valuable algorithm for large-scale machine learning, but has proven difficult to parallelize on conventional architectures because of communication and memory access issues. The HogWild series of mixed logically distributed and physically multi-threaded algorithms overcomes these issues for problems with sparse characteristics by using multiple local model vectors with asynchronous atomic updates. While this approach has proven effective for several reported examples, there are others, especially very sparse cases, that do not scale as well. This paper discusses an SGD Support Vector Machine (SVM) on a cacheless migrating thread architecture using the Hogwild algorithms as a framework. Our implementations on this novel architecture achieved superior hardware efficiency and scalability over that of a conventional cluster using MPI. Furthermore these improvements were gained using naive data partitioning techniques and hardware with substantially less compute capability than that present in conventional systems. 
    more » « less
  5. Communication compression has become a key strategy to speed up distributed optimization. However, existing decentralized algorithms with compression mainly focus on compressing DGD-type algorithms. They are unsatisfactory in terms of convergence rate, stability, and the capability to handle heterogeneous data. Motivated by primal-dual algorithms, this paper proposes the first \underline{L}in\underline{EA}r convergent \underline{D}ecentralized algorithm with compression, LEAD. Our theory describes the coupled dynamics of the inexact primal and dual update as well as compression error, and we provide the first consensus error bound in such settings without assuming bounded gradients. Experiments on convex problems validate our theoretical analysis, and empirical study on deep neural nets shows that LEAD is applicable to non-convex problems. 
    more » « less