Distributed Optimization Over Time-Varying Directed Graphs
The DOI matches its Crossref record; 35 of the 36 references it cites were found in a registry or an index; 2 of the 13 sources its claims cite were checked for support.
Abstract
We consider distributed optimization by a collection of nodes, each having access to its own convex function, whose collective goal is to minimize the sum of the functions. The communications between nodes are described by a time-varying sequence of directed graphs, which is uniformly strongly connected. For such communications, assuming that every node knows its out-degree, we develop a broadcast-based algorithm, termed the subgradient-push, which steers every node to an optimal value under a standard assumption of subgradient boundedness. The subgradient-push requires no knowledge of either the number of agents or the graph sequence to implement. Our analysis shows that the subgradient-push algorithm converges at a rate of O(\ln t √t). The proportionality constant in the convergence rate depends on the initial values at the nodes, the subgradient norms and, more interestingly, on both the speed of the network information diffusion and the imbalances of influence among the nodes.
Overview
CoChat Summary
TL;DR
The paper introduces the subgradient-push algorithm for distributed convex optimization over time-varying directed graphs that are uniformly strongly connected, requiring each node only to know its outdegree. It proves convergence to a common optimal solution and an objective convergence rate of \(O(\ln t/\sqrt{t})\).
Key findings
- The subgradient-push method drives all nodes to the same optimal point under convexity, nonempty optimal set, bounded subgradients, and standard decaying stepsize conditions.
- For stepsize \(\alpha(t)=1/\sqrt{t}\), the objective converges at rate \(O(\ln t/\sqrt{t})\), with the constant depending on initial values, subgradient norm bounds, the network diffusion speed \(\lambda\), and the influence-imbalance measure \(\delta\).
- For regular directed graphs, \(\delta=1\); in general, \(1/(\delta(1-\lambda))\) is bounded exponentially by \(n^{2nB}\) in the worst case, but for time-varying regular directed graphs it scales polynomially in \(n\).
- Simulations on a scalar quadratic estimation problem with \(n=1000\) nodes and threshold \(\|z(t)-\theta^*1\|\le 0.1\) show sub-geometric decay and time scaling linearly in the number of nodes from \(n=10\) to \(90\).
Methods
- Theoretical design and analysis of a broadcast-based distributed subgradient algorithm combining subgradient updates with the push-sum protocol over time-varying directed graphs.
- Convergence proofs rely on a perturbed push-sum protocol, consensus-product bounds, and deterministic supermartingale-type convergence lemmas; the rate proof uses \(\alpha(t)=1/\sqrt{t}\) and node-maintained running averages \(\tilde z_i(t)\).
- Simulations optimize \(F(\theta)=\sum_{i=1}^n p_i(\theta-u_i)^2\), with half the \(p_i\)'s set to zero and half uniformly random in \([0,1]\), Gaussian initial points, and graph sequences including random two-out-neighbor graphs and alternating undirected star graphs.
Limitations
- The paper states that Theorem 2 does not address convergence of the sequences \(\{\tilde z_i(t)\}\), and it lists open problems on how \(\lambda\) depends on graph-sequence properties and how \(\delta\) depends on combinatorial graph features such as diameters and smallest cuts.
- The text makes evident that the simulations are restricted to a scalar quadratic estimation problem and specific graph models, so empirical behavior for general convex functions or higher-dimensional problems is not demonstrated.
- The algorithm assumes every node knows its outdegree \(d_i(t)\) at every time \(t\), which is an implementation requirement rather than a limitation the paper frames as such.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
2 Valid · 1 Not valid · 3 Not checked · 36 Own results and interpretations
A result the paper states — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Here, we generalize this result by showing that the protocol remains convergent even if the state of the nodes is perturbed at each step, as long as the perturbations decay to zero.”
- Also stated in: § III. PERTURBED PUSH-SUM PROTOCOL · p. 5 · plain
“When the perturbations are nonzero, Lemma 1 states that if these perturbations decay to zero, then the push-sum method still converges.”
boostedstates a possible cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on 2 assumptions not shown here.
- Also stated in: § III. PERTURBED PUSH-SUM PROTOCOL · p. 5 · plain
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Indeed, Theorem 2 shows that for a class of time-varying regular directed graphs, 1 / ( δ (1 -λ )) scales polynomially in n .”
boostedstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption not shown here.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The proportionality constant in the convergence rate depends on the initial values at the nodes, the subgradient norms and, more interestingly, on both the speed of the network information diffusion and the imbalances of influence among the nodes.”
- Also stated in: § I. INTRODUCTION · p. 1 · plain
“Our results show that it converges at a rate of O ( ln t/ √ t ) , where the constant depends, among other factors, on the information diffusion speed of the corresponding directed graph sequence and a measure of the influence imbalance among the nodes.”
plainstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on 3 assumptions not shown here.
- Also stated in: § I. INTRODUCTION · p. 1 · plain
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Our analysis shows that the subgradientpush algorithm converges at a rate of O ( ln t/ √ t ) .”
- Also stated in: § B. Our results · p. 3 · boosted
“We then demonstrate that the subgradient-push converges at a rate of O (ln t/ √ t ) ; this is formally stated in the following theorem.”
boostedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on 5 assumptions not shown here.
- Also stated in: § B. Our results · p. 3 · boosted
A result the paper states — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The subgradient-push requires no knowledge of either the number of agents or the graph sequence to implement.”
- Also stated in: § I. INTRODUCTION · p. 1 · plain
“By contrast, our results work in time-varying networks and are fully distributed, requiring no knowledge of either the graph sequence or the number of agents.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption not shown here.
- Also stated in: § I. INTRODUCTION · p. 1 · plain
12 more claims
Create a free account to see all 42 claims and every source verdict.
- It may state two separate claims as one. 13
- Probably not a substantial claim. 8
- The quote only partly states this claim. 7
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The cited source says what the claim attributes to it.
In the citing paper
“Assumption 2 is typical for many results in multi-agent coordination and distributed optimization [42]. It is much weaker than the assumption of every G un ( k ) being connected.”
From Achieving Geometric Convergence for Distributed Optimization Over Time-Varying Graphs (2017) · cites this paper as [42]
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ B. Our results · p. 3
“Theorem 2: Suppose that all the assumptions of Theorem 1 hold, and let α ( t ) = 1 / √ t for t ≥ 1 .”
§ IV. CONVERGENCE RESULTS FOR SUBGRADIENT-PUSH METHOD · p. 12
“Also, by assumption we have ∑ ∞ t =1 α ( t ) = ∞ and ∑ ∞ t =1 α 2 ( t ) < ∞ .”
Citations
Cites
Coordination of groups of mobile autonomous agents using nearest neighbor rules
Jadbabaie, A., Lin, J., & Morse, A. S. (2003) · IEEE Transactions on Automatic Control · doi:10.1109/tac.2003.812781 · Cited by 8,339
Distributed Subgradient Methods for Multi-Agent Optimization
Nedic, A., & Ozdaglar, A. (2009) · IEEE Transactions on Automatic Control · doi:10.1109/tac.2008.2009515 · Cited by 3,719 · cited 2×
“Distributed optimization of a sum of convex functions has received a surge of interest in recent years [18], [23], [9], [14], [12], [13], [28], [16], [3], [24], [5].”
BackgroundMethodStability of multiagent systems with time-dependent communication links
Moreau, L. (2005) · IEEE Transactions on Automatic Control · doi:10.1109/tac.2004.841888 · Cited by 2,706
Distributed asynchronous deterministic and stochastic gradient optimization algorithms
Tsitsiklis, J., Bertsekas, D., & Athans, M. (1986) · IEEE Transactions on Automatic Control · doi:10.1109/tac.1986.1104412 · Cited by 1,967
Gossip-based computation of aggregate information
Kempe, D., Dobra, A., & Gehrke, J. (2004) · 44th Annual IEEE Symposium on Foundations of Computer Science, 2003. Proceedings. · doi:10.1109/sfcs.2003.1238221 · Cited by 1,480
Dual Averaging for Distributed Optimization: Convergence Analysis and Network Scaling
Duchi, J. C., Agarwal, A., & Wainwright, M. J. (2012) · IEEE Transactions on Automatic Control · doi:10.1109/tac.2011.2161027 · Cited by 947
Distributed optimization in sensor networks
Rabbat, M., & Nowak, R. (2004) · Proceedings of the 3rd international symposium on Information processing in sensor networks · doi:10.1145/984622.984626 · Cited by 901 · cited 1×
“For example, the problems including robust statistical inference [22], formation control [20], non-autonomous power control [25], distributed message routing [19], and spectrum access coordination [10], can be reduced to variations of this problem.”
Dual Averaging for Distributed Optimization: Convergence Analysis and Network Scaling
Duchi, J. C., Agarwal, A., & Wainwright, M. J. (2010) · Cited by 884
Convergence in Multiagent Coordination, Consensus, and Flocking
Blondel, V. D., Hendrickx, J. M., Olshevsky, A., & Tsitsiklis, J. N. (2006) · Proceedings of the 44th IEEE Conference on Decision and Control · doi:10.1109/cdc.2005.1582620 · Cited by 859
Self-citationDistributed Stochastic Subgradient Projection Algorithms for Convex Optimization
Sundhar Ram, S., Nedić, A., & Veeravalli, V. V. (2010) · Journal of Optimization Theory and Applications · doi:10.1007/s10957-010-9737-7 · Cited by 814 · cited 2×
“Specifically, it is standard for a distributed subgradient method to converge at the rate of O(ln t/ √ t) with the constant depending on the subgradientnorm upper bounds Li and the initial conditions xi(0) [5], [24].”
Self-citation
Research works printed but not linked: 2
- [29] K.I. Tsianos. The role of the Network in Distributed Optimization Algorithms: Convergence Rates, Scalability, Communication / Computation Tradeoffs and Communication Delays . PhD thesis, McGill University, Dept. of Electrical and Computer Engineering, 2013. -
- [7] F. Iutzeler, P. Ciblat, and W. Hachem. Analysis of sum-weightlike algorithms for averaging inwireless sensor networks. IEEE Transactions on Signal Processing , 6(11), 2013. -
Cited by
Items 1,201–1,226 of 1,226
Hybrid Decentralized Task Allocation for Swarm Rover Search Using Silent Auctions and CBBA
Us, K. S., Gupta, Y., Ravi, D. T., et al. (2026) · doi:10.1109/iecct68664.2026.11541702
“Co-operative control strategies and distributed consensus methods are therefore important research fields for the coordination of scalable swarms in constrained operational environments [2]– [4].”
Liu, H., Cheng, M., Yu, W., & Zheng, W. X. (2026) · Science China Information Sciences · doi:10.1007/s11432-025-4887-6
Jiang, L., Ming, X., Leng, J., Xu, J., Sha, W., & Zhang, X. (2026) · Advanced Engineering Informatics · doi:10.1016/j.aei.2026.104866 · Cited by 2
Li, Z., Ran, L., Li, J., & Zheng, L. (2026) · Mathematical and Computational Applications · doi:10.3390/mca31030091
“Notably, the works [14–22] focus on optimization of a single objective function, which is commonly formulated as min ∑ ni = 1 f i ( x ) .”
InfluentialIntertwined Escort Dynamics: A Dynamical Framework for Constrained OptimizationPreprint
Suryawanshi, U. S., Arekar, S., Kumar, S., et al. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.6876570
Zhao, J., Tian, L., Liu, H., et al. (2026) · IEEE Transactions on Control of Network Systems · doi:10.1109/tcns.2026.3702287
“Lemma 4 (See [39]): Under Assumption (1), for (10b) and (10d), there exist constants 0 γ 1 , γ 2 ∞ and 0 λ 1 , λ 2 1 such that As a direct consequence of Lemma 4, the following lemma gives the linear convergence of [ Z k ] − 1 and [ E ki ] − 1 , which will be used in the subsequent parts.”
Tian, L., Liu, H., Wang, H., & Yu, W. (2026) · IEEE Transactions on Control of Network Systems · doi:10.1109/tcns.2026.3704507
“In many practical applications, the communication graphs among agents are inherently unbalanced, such as networks with asymmetric communication links or heterogeneous sensor nodes with nonuniform communication ranges [17].”
InfluentialZhu, Y., Zhang, G., Liu, H., & Li, T. (2026) · Journal of the Franklin Institute · doi:10.1016/j.jfranklin.2026.108829
Liu, J., Wang, H., Cai, S., et al. (2026) · Chaos Solitons & Fractals · doi:10.1016/j.chaos.2026.118630
Distributed Online Optimization with Differential Privacy and Quantized Communication
Dai, M., & Fang, Y. (2026) · Chinese Control and Decision Conference (CCDC) · doi:10.1109/ccdc69976.2026.11560824
Mixing Matrix Design for Energy-efficient Decentralized Federated Learning
Zhang, X., Nguyen, T., & He, T. (2026) · doi:10.1109/infocom59046.2026.11571245
Du, Y., Li, J., Zhang, B., et al. (2026) · IEEE Transactions on Aerospace and Electronic Systems · doi:10.1109/taes.2026.3708337
Cheng, M., Liu, H., Tian, L., et al. (2026) · IEEE Transactions on Control of Network Systems · doi:10.1109/tcns.2026.3710315
“When only one cluster is involved, it becomes a distributed optimization problem with intracluster consensus, as studied in [21], [22], and [23].”
InfluentialDistributed Adaptive Gradient Tracking for Stochastic Nonconvex Optimization Over Directed Networks
Guan, M., He, C., & Zhang, K. (2026) · IEEE Internet of Things Journal · doi:10.1109/jiot.2026.3711154
Cooperative Online Safe Reinforcement Learning over Multiagent Networks
Ikeda, K., Hayashi, N., & Inuiguchi, M. (2026) · 2006 14th Mediterranean Conference on Control and Automation · doi:10.1109/med70602.2026.11598541
Learning-Based Online Cooperative Control over Time-Varying Directed Networks
ISHIKAWA, D., Hayashi, N., & Inuiguchi, M. (2026) · 2006 14th Mediterranean Conference on Control and Automation · doi:10.1109/med70602.2026.11598080
A Distributed Aggregative Optimization Algorithm for Unbalanced Directed Graphs
Chen, Y., & Yang, Q. (2026) · Journal of Scientific Computing · doi:10.1007/s10915-026-03401-w
Das, S., Schenato, L., & Dey, S. (2026) · European Journal of Control · doi:10.1016/j.ejcon.2026.101610
Wang, X., Chang, S., Dai, M., Yu, H., Tu, Y., Zhu, H., & Li, B. (2026) · IEEE International Conference on Distributed Computing Systems (ICDCS) · doi:10.1109/2575-8411.2026.00089
Mobility-Assisted Decentralized Federated Learning: Convergence Analysis and a Data-Driven Approach
Jahani, R., Reza, M. F., & Dai, H. (2026) · IEEE Transactions on Cognitive Communications and Networking · doi:10.1109/tccn.2026.3720485
Zhu, Y., Niu, B., Wang, Z., Song, X., Zhao, X., & Niu, Y. (2026) · Journal of the Franklin Institute · doi:10.1016/j.jfranklin.2026.109044
Distributed Adam for Nonconvex Optimization on Time-Varying Directed Networks
ISHIKAWA, D., Hayashi, N., & Inuiguchi, M. (2026) · IEEE Access · doi:10.1109/access.2026.3732357
Intertwined escort dynamics: A dynamical framework for constrained optimization
Suryawanshi, U. S., Arekar, S., Kumar, S., Pimple, B. B., & Wagh, S. R. (2026) · Electric Power Systems Research · doi:10.1016/j.epsr.2026.114229
Wang, P., Liu, L., & Feng, G. (2026) · Automatica · doi:10.1016/j.automatica.2026.113296
Fu, Y., & Mo, L. (2026) · Journal of Control and Decision · doi:10.1080/23307706.2026.2735344
A Two‐Average‐Tracking Scheme in Distributed Optimization With a Small‐Gain Convergence Analysis
Xiao, Y., & Ni, W. (2026) · Asian Journal of Control · doi:10.1002/asjc.70243
Metadata
- Authors
- Nedic, A., & Olshevsky, A.
- Affiliations
- University of Illinois Urbana-Champaign (US)
- ORCID
- Olshevsky, Alex
- Venue
- IEEE Transactions on Automatic Control
- Published
- 2015-03
- Volume
- 60
- Issue
- 3
- Pages
- 601-615
- ISSN
- 0018-9286 (ISSN-L)
- 1558-2523
- OpenAlex
- W2137435346
- Cited by
- 1,187
- References
- 33
- Citations by year
- 2026: 67 · 2025: 114 · 2024: 108 · 2023: 125 · 2022: 115 · 2021: 103 · 2020: 138 · 2019: 139 · 2018: 94 · 2017: 89 · 2016: 61 · 2015: 24
- FWCI
- 69.00
- Type
- journal article
- Language
- en
- Open access
- closed
- Locations
- IEEE Transactions on Automatic Control (publishedVersion)
- Funders
- National Science Foundation (NSF) (CMMI 07-42538, CCF 11-11342)
- ONR Navy Basic Research Challenge (N00014-12-1-0998)
- Topics
- Physical Sciences › Engineering › Control and Systems Engineering › Distributed Control Multi-Agent Systems
- Physical Sciences › Computer Science › Computational Theory and Mathematics › Optimization and Variational Analysis
- Physical Sciences › Engineering › Control and Systems Engineering › Advanced Control Systems Optimization
- Keywords
- distributed optimization, time-varying directed graphs, convex optimization, convergence rate, multi-agent systems
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.