Search Results

Documents authored by Anagnostides, Ioannis

Document
DOI: 10.4230/LIPIcs.ITCS.2024.5
On the Complexity of Computing Sparse Equilibria and Lower Bounds for No-Regret Learning in Games

Authors: Ioannis Anagnostides, Alkis Kalavasis, Tuomas Sandholm, and Manolis Zampetakis

Published in: LIPIcs, Volume 287, 15th Innovations in Theoretical Computer Science Conference (ITCS 2024)

Abstract
Characterizing the performance of no-regret dynamics in multi-player games is a foundational problem at the interface of online learning and game theory. Recent results have revealed that when all players adopt specific learning algorithms, it is possible to improve exponentially over what is predicted by the overly pessimistic no-regret framework in the traditional adversarial regime, thereby leading to faster convergence to the set of coarse correlated equilibria (CCE) - a standard game-theoretic equilibrium concept. Yet, despite considerable recent progress, the fundamental complexity barriers for learning in normal- and extensive-form games are poorly understood. In this paper, we make a step towards closing this gap by first showing that - barring major complexity breakthroughs - any polynomial-time learning algorithms in extensive-form games need at least 2^{log^{1/2 - o(1)} |𝒯|} iterations for the average regret to reach below even an absolute constant, where |𝒯| is the number of nodes in the game. This establishes a superpolynomial separation between no-regret learning in normal- and extensive-form games, as in the former class a logarithmic number of iterations suffices to achieve constant average regret. Furthermore, our results imply that algorithms such as multiplicative weights update, as well as its optimistic counterpart, require at least 2^{(log log m)^{1/2 - o(1)}} iterations to attain an O(1)-CCE in m-action normal-form games under any parameterization. These are the first non-trivial - and dimension-dependent - lower bounds in that setting for the most well-studied algorithms in the literature. From a technical standpoint, we follow a beautiful connection recently made by Foster, Golowich, and Kakade (ICML '23) between sparse CCE and Nash equilibria in the context of Markov games. Consequently, our lower bounds rule out polynomial-time algorithms well beyond the traditional online learning framework, capturing techniques commonly used for accelerating centralized equilibrium computation.
Cite as

Ioannis Anagnostides, Alkis Kalavasis, Tuomas Sandholm, and Manolis Zampetakis. On the Complexity of Computing Sparse Equilibria and Lower Bounds for No-Regret Learning in Games. In 15th Innovations in Theoretical Computer Science Conference (ITCS 2024). Leibniz International Proceedings in Informatics (LIPIcs), Volume 287, pp. 5:1-5:24, Schloss Dagstuhl – Leibniz-Zentrum für Informatik (2024)

Copy BibTex To Clipboard

@InProceedings{anagnostides_et_al:LIPIcs.ITCS.2024.5,
  author =	{Anagnostides, Ioannis and Kalavasis, Alkis and Sandholm, Tuomas and Zampetakis, Manolis},
  title =	{{On the Complexity of Computing Sparse Equilibria and Lower Bounds for No-Regret Learning in Games}},
  booktitle =	{15th Innovations in Theoretical Computer Science Conference (ITCS 2024)},
  pages =	{5:1--5:24},
  series =	{Leibniz International Proceedings in Informatics (LIPIcs)},
  ISBN =	{978-3-95977-309-6},
  ISSN =	{1868-8969},
  year =	{2024},
  volume =	{287},
  editor =	{Guruswami, Venkatesan},
  publisher =	{Schloss Dagstuhl -- Leibniz-Zentrum f{\"u}r Informatik},
  address =	{Dagstuhl, Germany},
  URL =		{https://drops.dagstuhl.de/entities/document/10.4230/LIPIcs.ITCS.2024.5},
  URN =		{urn:nbn:de:0030-drops-195334},
  doi =		{10.4230/LIPIcs.ITCS.2024.5},
  annote =	{Keywords: No-regret learning, extensive-form games, multiplicative weights update, optimism, lower bounds}
}
Document
DOI: 10.4230/LIPIcs.DISC.2022.6
Almost Universally Optimal Distributed Laplacian Solvers via Low-Congestion Shortcuts

Authors: Ioannis Anagnostides, Christoph Lenzen, Bernhard Haeupler, Goran Zuzic, and Themis Gouleakis

Published in: LIPIcs, Volume 246, 36th International Symposium on Distributed Computing (DISC 2022)

Abstract
In this paper, we refine the (almost) existentially optimal distributed Laplacian solver recently developed by Forster, Goranci, Liu, Peng, Sun, and Ye (FOCS `21) into an (almost) universally optimal distributed Laplacian solver. Specifically, when the topology is known (i.e., the Supported-CONGEST model), we show that any Laplacian system on an n-node graph with shortcut quality SQ(G) can be solved after n^{o(1)} SQ(G) log(1/ε) rounds, where ε is the required accuracy. This almost matches our lower bound that guarantees that any correct algorithm on G requires Ω̃(SQ(G)) rounds, even for a crude solution with ε ≤ 1/2. Several important implications hold in the unknown-topology (i.e., standard CONGEST) case: for excluded-minor graphs we get an almost universally optimal algorithm that terminates in D ⋅ n^{o(1)} log(1/ε) rounds, where D is the hop-diameter of the network; as well as n^{o(1)} log (1/ε)-round algorithms for the case of SQ(G) ≤ n^{o(1)}, which holds for most networks of interest. Conditioned on improvements in state-of-the-art constructions of low-congestion shortcuts, the CONGEST results will match the Supported-CONGEST ones. Moreover, following a recent line of work in distributed algorithms, we consider a hybrid communication model which enhances CONGEST with limited global power in the form of the node-capacitated clique (NCC) model. In this model, we show the existence of a Laplacian solver with round complexity n^{o(1)} log(1/ε). The unifying thread of these results, and our main technical contribution, is the study of a novel ρ-congested generalization of the standard part-wise aggregation problem. We develop near-optimal algorithms for this primitive in the Supported-CONGEST model, almost-optimal algorithms in (standard) CONGEST (with the additional overhead due to standard barriers), as well as a simple algorithm for bounded-treewidth graphs with a quadratic dependence on the congestion ρ. This primitive can be readily used to accelerate the Laplacian solver of Forster, Goranci, Liu, Peng, Sun, and Ye, and we believe it will find further independent applications in the future.
Cite as

Ioannis Anagnostides, Christoph Lenzen, Bernhard Haeupler, Goran Zuzic, and Themis Gouleakis. Almost Universally Optimal Distributed Laplacian Solvers via Low-Congestion Shortcuts. In 36th International Symposium on Distributed Computing (DISC 2022). Leibniz International Proceedings in Informatics (LIPIcs), Volume 246, pp. 6:1-6:20, Schloss Dagstuhl – Leibniz-Zentrum für Informatik (2022)

Copy BibTex To Clipboard

@InProceedings{anagnostides_et_al:LIPIcs.DISC.2022.6,
  author =	{Anagnostides, Ioannis and Lenzen, Christoph and Haeupler, Bernhard and Zuzic, Goran and Gouleakis, Themis},
  title =	{{Almost Universally Optimal Distributed Laplacian Solvers via Low-Congestion Shortcuts}},
  booktitle =	{36th International Symposium on Distributed Computing (DISC 2022)},
  pages =	{6:1--6:20},
  series =	{Leibniz International Proceedings in Informatics (LIPIcs)},
  ISBN =	{978-3-95977-255-6},
  ISSN =	{1868-8969},
  year =	{2022},
  volume =	{246},
  editor =	{Scheideler, Christian},
  publisher =	{Schloss Dagstuhl -- Leibniz-Zentrum f{\"u}r Informatik},
  address =	{Dagstuhl, Germany},
  URL =		{https://drops.dagstuhl.de/entities/document/10.4230/LIPIcs.DISC.2022.6},
  URN =		{urn:nbn:de:0030-drops-171978},
  doi =		{10.4230/LIPIcs.DISC.2022.6},
  annote =	{Keywords: Distributed algorithms, Laplacian solvers, low-congestion shortcuts}
}
Document
DOI: 10.4230/LIPIcs.DISC.2021.5
Deterministic Distributed Algorithms and Lower Bounds in the Hybrid Model

Authors: Ioannis Anagnostides and Themis Gouleakis

Published in: LIPIcs, Volume 209, 35th International Symposium on Distributed Computing (DISC 2021)

Abstract
The HYBRID model was recently introduced by Augustine et al. [John Augustine et al., 2020] in order to characterize from an algorithmic standpoint the capabilities of networks which combine multiple communication modes. Concretely, it is assumed that the standard LOCAL model of distributed computing is enhanced with the feature of all-to-all communication, but with very limited bandwidth, captured by the node-capacitated clique (NCC). In this work we provide several new insights on the power of hybrid networks for fundamental problems in distributed algorithms. First, we present a deterministic algorithm which solves any problem on a sparse n-node graph in 𝒪̃(√n) rounds of HYBRID, where the notation 𝒪̃(⋅) suppresses polylogarithmic factors of n. We combine this primitive with several sparsification techniques to obtain efficient distributed algorithms for general graphs. Most notably, for the all-pairs shortest paths problem we give deterministic (1 + ε)- and log n/log log n-approximate algorithms for unweighted and weighted graphs respectively with round complexity 𝒪̃(√n) in HYBRID, closely matching the performance of the state of the art randomized algorithm of Kuhn and Schneider [Kuhn and Schneider, 2020]. Moreover, we make a connection with the Ghaffari-Haeupler framework of low-congestion shortcuts [Mohsen Ghaffari and Bernhard Haeupler, 2016], leading - among others - to a (1 + ε)-approximate algorithm for Min-Cut after 𝒪(polylog (n)) rounds, with high probability, even if we restrict local edges to transfer 𝒪(log n) bits per round. Finally, we prove via a reduction from the set disjointness problem that Ω̃(n^{1/3}) rounds are required to determine the radius of an unweighted graph, as well as a (3/2 - ε)-approximation for weighted graphs. As a byproduct, we show an Ω̃(n) round-complexity lower bound for computing a (4/3 - ε)-approximation of the radius in the broadcast variant of the congested clique, even for unweighted graphs.
Cite as

Ioannis Anagnostides and Themis Gouleakis. Deterministic Distributed Algorithms and Lower Bounds in the Hybrid Model. In 35th International Symposium on Distributed Computing (DISC 2021). Leibniz International Proceedings in Informatics (LIPIcs), Volume 209, pp. 5:1-5:19, Schloss Dagstuhl – Leibniz-Zentrum für Informatik (2021)

Copy BibTex To Clipboard

@InProceedings{anagnostides_et_al:LIPIcs.DISC.2021.5,
  author =	{Anagnostides, Ioannis and Gouleakis, Themis},
  title =	{{Deterministic Distributed Algorithms and Lower Bounds in the Hybrid Model}},
  booktitle =	{35th International Symposium on Distributed Computing (DISC 2021)},
  pages =	{5:1--5:19},
  series =	{Leibniz International Proceedings in Informatics (LIPIcs)},
  ISBN =	{978-3-95977-210-5},
  ISSN =	{1868-8969},
  year =	{2021},
  volume =	{209},
  editor =	{Gilbert, Seth},
  publisher =	{Schloss Dagstuhl -- Leibniz-Zentrum f{\"u}r Informatik},
  address =	{Dagstuhl, Germany},
  URL =		{https://drops.dagstuhl.de/entities/document/10.4230/LIPIcs.DISC.2021.5},
  URN =		{urn:nbn:de:0030-drops-148077},
  doi =		{10.4230/LIPIcs.DISC.2021.5},
  annote =	{Keywords: Distributed Computing, Hybrid Model, Sparse Graphs, Deterministic Algorithms, All-Pairs Shortest Paths, Minimum Cut, Radius}
}
Questions / Remarks / Feedback
X

Feedback for Dagstuhl Publishing


Thanks for your feedback!

Feedback submitted

Could not send message

Please try again later or send an E-mail
© 2023-2024 Schloss Dagstuhl – LZI GmbH Imprint Privacy Contact