[Paper Review] Exponential Convergence of the Discrete-Time Altafini Model

Ji Liu, Xudong Chen+2 more|arXiv (Cornell University)|Dec 22, 2015

Opinion Dynamics and Social Influence32 references3 citations

TL;DR

This paper establishes necessary and sufficient conditions for exponential convergence in the discrete-time Altafini model of signed opinion dynamics using a graphical approach. It proves that under repeatedly jointly strong connectivity, the system exponentially converges to a two-clustering state if and only if the signed digraph sequence is repeatedly jointly structurally balanced for that clustering, and to zero exponentially fast if and only if the sequence is repeatedly jointly structurally unbalanced, with an explicit upper bound on the convergence rate.

ABSTRACT

This paper considers the discrete-time version of Altafini's model for opinion dynamics in which the interaction among a group of agents is described by a time-varying signed digraph. Prompted by an idea from [1], exponential convergence of the system is studied using a graphical approach. Necessary and sufficient conditions for exponential convergence with respect to each possible type of limit states are provided. Specifically, under the assumption of repeatedly jointly strong connectivity, it is shown that (1) a certain type of two-clustering will be reached exponentially fast for almost all initial conditions if, and only if, the sequence of signed digraphs is repeatedly jointly structurally balanced corresponding to that type of two-clustering; (2) the system will converge to zero exponentially fast for all initial conditions if, and only if, the sequence of signed digraphs is repeatedly jointly structurally unbalanced. An upper bound on the convergence rate is also provided.

Motivation & Objective

• To determine necessary and sufficient conditions for exponential convergence in the discrete-time Altafini model under time-varying signed digraphs.
• To characterize the convergence rate and provide an upper bound on the rate of convergence.
• To relax the strong connectivity assumption in the time-invariant case, allowing for less restrictive connectivity conditions.
• To extend prior work by providing full proofs, convergence rate analysis, and treatment of non-symmetric interaction and sign structures.

Proposed method

• A lifting approach inspired by Hendrickx (2014) is employed to relate the Altafini model to an expanded DeGroot consensus model with special structure.
• Graph-theoretic tools are used to analyze the signed digraphs, particularly focusing on mutual reachability and rootedness.
• The spectral properties of the system matrix are analyzed using Gersgorin circle theorem and the inequality ρ(M) ≤ ρ(|M|) for matrix spectral radius.
• The analysis distinguishes between structurally balanced and unbalanced digraphs, linking them to convergence to two-clustering or zero states.
• A permutation-based matrix decomposition is used to isolate the submatrix corresponding to the set of roots, enabling eigenvalue analysis.
• The convergence rate upper bound is derived from the spectral properties of the absolute value matrix and the structure of the signed graph.

Experimental results

Research questions

• RQ1 What are the necessary and sufficient conditions on time-varying signed digraphs for exponential convergence to a specific two-clustering state in the discrete-time Altafini model?
• RQ2 Under what conditions does the system converge exponentially fast to zero for all initial conditions?
• RQ3 How can the convergence rate be quantitatively bounded in terms of the network structure and dynamics?
• RQ4 Can the strong connectivity assumption be relaxed in the time-invariant case while preserving exponential convergence?
• RQ5 How does the asymmetry in interaction and sign structure affect convergence behavior compared to symmetric cases?

Key findings

• Exponential convergence to a specific two-clustering state occurs if and only if the sequence of signed digraphs is repeatedly jointly structurally balanced with respect to that clustering.
• Exponential convergence to zero occurs for all initial conditions if and only if the sequence of signed digraphs is repeatedly jointly structurally unbalanced.
• Under repeatedly jointly strong connectivity, the system always reaches modulus consensus, with two-clustering and zero consensus as special cases.
• An upper bound on the exponential convergence rate is provided, derived from the spectral properties of the system matrix and its absolute value.
• For time-invariant systems, exponential convergence is established under a less restrictive connectivity condition than strong connectivity, provided the root subgraph structure satisfies specific balance conditions.
• The eigenvalue structure of the system matrix is fully characterized: when the root subgraph is structurally balanced, there is an eigenvalue at 1; otherwise, all eigenvalues are strictly inside the unit circle.