ResearchTrend.AI
  • Papers
  • Communities
  • Events
  • Blog
  • Pricing
Papers
Communities
Social Events
Terms and Conditions
Pricing
Parameter LabParameter LabTwitterGitHubLinkedInBlueskyYoutube

© 2025 ResearchTrend.AI, All rights reserved.

  1. Home
  2. Papers
  3. 1506.04681
101
37
v1v2 (latest)

Byzantine Multi-Agent Optimization: Part I

15 June 2015
Lili Su
Nitin H. Vaidya
ArXiv (abs)PDFHTML
Abstract

We study Byzantine fault-tolerant distributed optimization of a sum of convex (cost) functions with real-valued scalar input/ouput. In particular, the goal is to optimize a global cost function 1∣N∣∑i∈Nhi(x)\frac{1}{|\mathcal{N}|}\sum_{i\in \mathcal{N}} h_i(x)∣N∣1​∑i∈N​hi​(x), where N\mathcal{N}N is the set of non-faulty agents, and hi(x)h_i(x)hi​(x) is agent iii's local cost function, which is initially known only to agent iii. In general, when some of the agents may be Byzantine faulty, the above goal is unachievable, because the identity of the faulty agents is not necessarily known to the non-faulty agents, and the faulty agents may behave arbitrarily. Since the above global cost function cannot be optimized exactly in presence of Byzantine agents, we define a weaker version of the problem. The goal for the weaker problem is to generate an output that is an optimum of a function formed as a convex combination of local cost functions of the non-faulty agents. More precisely, for some choice of weights αi\alpha_iαi​ for i∈Ni\in \mathcal{N}i∈N such that αi≥0\alpha_i\geq 0αi​≥0 and ∑i∈Nαi=1\sum_{i\in \mathcal{N}}\alpha_i=1∑i∈N​αi​=1, the output must be an optimum of the cost function ∑i∈Nαihi(x)\sum_{i\in \mathcal{N}} \alpha_ih_i(x)∑i∈N​αi​hi​(x). Ideally, we would like αi=1∣N∣\alpha_i=\frac{1}{|\mathcal{N}|}αi​=∣N∣1​ for all i∈Ni\in \mathcal{N}i∈N -- however, this cannot be guaranteed due to the presence of faulty agents. In fact, we show that the maximum achievable number of nonzero weights (αi\alpha_iαi​'s) is ∣N∣−f|\mathcal{N}|-f∣N∣−f, where fff is the upper bound on the number of Byzantine agents. In addition, we present algorithms that ensure that at least ∣N∣−f|\mathcal{N}|-f∣N∣−f agents have weights that are bounded away from 0. We also propose a low-complexity suboptimal algorithm, which ensures that at least ⌈n2⌉−ϕ\lceil \frac{n}{2}\rceil-\phi⌈2n​⌉−ϕ agents have weights that are bounded away from 0, where nnn is the total number of agents, and ϕ\phiϕ (ϕ≤f\phi\le fϕ≤f) is the actual number of Byzantine agents.

View on arXiv
Comments on this paper