Group-Based Alternating Direction Method of Multipliers for Distributed Linear Classification

Wang, Huihui; Gao, Yang; Shi, Yinghuan; Wang, Ruili

doi:10.1109/tcyb.2016.2570808

Cited by 40 publications

(11 citation statements)

References 30 publications

(44 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The practices of social commerce have been evolving over time (Curty & Zhang, 2011) with the advent and growth of new technology, platforms, algorithms (Chen, Wang, Zhang, Wang, & Xu, 2019;Guo, Yi, Wang, Ye, & Zhao, 2014;Wang, Gao, Shi, & Wang, 2016;Yu, Gao, Wang, & Wang, 2018;Yu, Jung, Kim, & Jung, 2018) and social media tools. The social commerce venues have evolved from originally text-oriented blogs, social networking sites (SNSs) (Wang & Zhang, 2012) and social shopping websites, to today's most popular live streaming which combines text, image, audio, and video, and provides live interactions.…”

Section: Social Commercementioning

confidence: 99%

Discovery of behavioral patterns in online social commerce practice

Jia

Wang

Liu

2021

WIREs Data Min & Knowl

Self Cite

View full text Add to dashboard Cite

Discovery of behavioral patterns in online social commerce practice becomes important in this digital era. In this article, we propose a systematic approach to behavioral pattern discovery, and apply it in an emerging online social commerce venue: live streaming. We investigate behavioral patterns in gifting encouragement in live streaming to understand online social commerce practice. Our proposed approach is based on multiple triangulation, including data source triangulation (i.e., streamers, viewers, and actual behavior) and data collection method triangulation (i.e., interviews, focus groups, and observations). Through multiple triangulation, four behavioral patterns of gifting encouragement are discovered: (i) requesting a certain gift for providing a particular service, (ii) creating a raffle, (iii) eliciting competition between individuals, and (iv) eliciting competition between groups. This research reveals the special behavioral patterns in live streaming, and thus increases our knowledge of social commerce practices. This research provides a systematic approach to discover online behavioral patterns, and provides practical implications in live streaming platforms, especially in marketing and platform design.

show abstract

Section: Social Commercementioning

confidence: 99%

Discovery of behavioral patterns in online social commerce practice

Jia

Wang

Liu

2021

WIREs Data Min & Knowl

Self Cite

View full text Add to dashboard Cite

show abstract

“…The enhanced sparsity and fast convergence of iterative reweighted ADMM algorithm was demonstrated and validated. In Wang et al, the group‐based ADMM was proposed for distributed linear classification with large‐scale dataset, which is a crucial research topic in machine learning. In Hu et al, the proposed ADMM‐based algorithm can improve the reconstruction efficiency of parameter assessment by retrieval from signal encoding via simplifying the original complicated cost function into a composition of simpler optimization procedures.…”

Section: Introductionmentioning

confidence: 99%

Capacity reservation optimization for distributed satellite cluster networks

Zheng

Zhao

et al. 2018

Satell Commun Network

View full text Add to dashboard Cite

We investigate the problem of link capacity reservation in distributed satellite cluster networks. In the optimization problem, the objective is to minimize the link reservation cost under the link capacity constraints. The original optimization problem can be formulated as a set of the linear programs. For the small-scale networks, the general linear program solvers have the capability to solve the problem within an acceptable time. To solve the problem in the large-scale network, we propose a heuristic method on the basis of the alternating direction method of multipliers, in which the original problem is decoupled to parallel small problems that can be solved simply.In the heuristic, the upper and lower bounds are computed iteratively. When the upper and lower bounds are close enough, then the optimal value can be obtained approximately. The simulation results show that the proposed method is comparatively exact when the stopping criterion is properly chosen. KEYWORDSalternating direction method of multipliers, convex optimization, distributed satellite cluster, heuristic algorithm | INTRODUCTIONA distributed satellite cluster (DSC) 1-4 is a special constellation architecture that consists of numbers of interconnected satellites distributed on the adjacent orbital positions within given distance bounds. A DSC can be treated as a single satellite with more payload and stronger ability. In general, via short intersatellite link (about 100 km), the satellites in a DSC form a relatively stable network and cooperate to achieve some particular space-based missions. By long-distance intersatellite link (more than 10 000 km), a number of interconnecting DSCs located at geostationary Earth orbits can almost cover the surface of the Earth, thus forming a global backbone network. 5,6 Unlike the terrestrial network, intersatellite links in satellite networks are expensive because of the limitation of on-board payload and energy reserves. In a DSC, there is a collection of traffic demands, which specify the amount of data flows to be routed from some sources to some sinks.The network must consume link resources to provide enough link capacity for these demands. Because of the varied quality-of-service requirement and limited on-board resource, dynamic link capacity deployment in the network is a complex process, especially when the network serves a number of heterogeneous space missions. Therefore, a link capacity deployment plan, capacity reservation, can be developed according to the mission distribution, to provide strategy flexibility and resource use in a resource-constrained system. Hence, we address the capacity reservation problem 7 to find the optimal strategy in a DSC network. The capacity reservation formulates the transmission strategy according to the mission distribution in the network and then reserves the capacity of all the links optimally. In the optimization problem, we seek to minimize the sum of link costs subject to provide sufficient link capacities to meet the requirements of multiple missions. Ther...

show abstract

“…Some researchers have recently made some improvements to ADMM. In terms of the network structure, to make the communication more efficient, distributed ADMM (D‐ADMM) and group‐based ADMM (GADMM) were proposed in the works of Mota et al 39 and Wang et al, 40 respectively. In contrast, in terms of distributed data, Wang et al 41 considered the imbalance of big data while using distributed learning with ADMM.…”

Section: Introductionmentioning

confidence: 99%

Consensus‐based distributed learning for robust convex optimization with a scenario approach

Cao

Feng

2018

Concurrency and Computation

View full text Add to dashboard Cite

This paper aims to solve the robust convex optimization (RCO) problem, where the constraints of RCO are parameterized with uncertainties, and the scenario approach is applied to transform RCO into standard convex optimization with a finite number of constraints through probabilistic approximation. The transformed problem is called a scenario problem (SP). Two consensus-based distributed learning algorithms for SP are designed in consideration of a large number of sampled constraints. One is based on the distributed average consensus (DAC), and the other is based on the alternating direction method of multipliers (ADMM). It has regulated that data distributed to nodes are not allowed to communicate. Simulation results indicate that the proposed algorithms are suitable for handling large-scale data and achieve excellent performance, with the ADMM-based algorithm performing the best. Furthermore, the DAC-based algorithm has certain advantages in terms of computational time and complexity. In addition, to improve the communicative efficiency based on a DAC, an efficient distributed average consensus (EDAC) is put forward. The average time for every node when using an EDAC is less than that of a DAC, despite the exact same performance. KEYWORDS alternating direction method of multipliers, consensus computation, distributed learning, robust convex optimization 1 INTRODUCTION 1.1 Background The so-called robust optimization (RO) problem indicates that the constraints are parameterized with uncertainties. The aim of this problem is to find a robust solution, satisfying all constraints for all possible situations and optimizing the objective function with the worst case, ie, the solution is not affected by the data. This problem has attracted significant interest 1-3 and has been widely applied in many areas, such as economic management, 4-6 planning and scheduling systems, 7,8 and communication systems, 9,10 among others. 2,3,11 In fact, RO is a semi-infinite optimization problem and cannot be efficiently solved numerically. In addition, some traditional methods for solving the optimization problem are no longer applicable. Therefore, it may cause that the optimal solution becomes feasible even if infeasible under the affection of the uncertainties. 12 Robust convex optimization (RCO) is a special case of RO with a convex objective function. The primary method for solving RCO is transforming it into an equivalent deterministic tractable convex problem under some mild assumptions, 3,13,14 and thus, the transformed optimization can be solved using existing convex optimization methods introduced in the work of Boyd and Vandenberghe. 15 However, a drawback of this primary approach is that the transformed convex optimization is often not as scalable as the nominal optimization. To solve RCO, a method called a scenario approach was studied in the works of Calafiore and Campi, 16 Campi and Garatti, 17 and De Farias and Van Roy, 18 where it was shown that most of the constraints of the original optimization are satisfied if the ...

show abstract

Group-Based Alternating Direction Method of Multipliers for Distributed Linear Classification

Cited by 40 publications

References 30 publications

Discovery of behavioral patterns in online social commerce practice

Discovery of behavioral patterns in online social commerce practice

Capacity reservation optimization for distributed satellite cluster networks

Consensus‐based distributed learning for robust convex optimization with a scenario approach

Contact Info

Product

Resources

About