Fastest Distributed Consensus on Petal Networks

Providing an analytical solution for the problem of finding Fastest Distributed Consensus (FDC) is one of the challenging problems in the field of sensor networks. Here in this work we present analytical solution for the problem of fastest distributed consensus averaging algorithm by means of stratification and semi-definite programming, for two particular types of Petal networks, namely symmetric and Complete Cored Symmetric (CCS) Petal networks. Our method in this paper is based on convexity of fastest distributed consensus averaging problem, and inductive comparing of the characteristic polynomials initiated by slackness conditions in order to find the optimal weights. Also certain types of leaves are introduced along with their optimal weights which are not achievable by the method used in this work if these leaves are considered individually.

💡 Research Summary

The paper tackles the “Fastest Distributed Consensus” (FDC) problem, which seeks the set of edge weights that yields the quickest convergence of a distributed averaging algorithm on a given graph. While the FDC problem is known to be convex and can be expressed as a semi‑definite program (SDP), closed‑form solutions have been limited to very simple topologies. The authors focus on two families of graphs that they call “petal networks”: (1) symmetric petal networks, where a single central node is connected to several identical leaf sub‑graphs (the petals), and (2) Complete‑Cored Symmetric (CCS) petal networks, where the core itself is a complete graph and each core node attaches to identical petals.

The methodological core of the work is a combination of stratification (layering) and SDP. By exploiting the inherent symmetry of the petal structures, the authors partition the node set into strata such that all nodes within a stratum are interchangeable. This reduces the original SDP, which would involve an N‑by‑N weight matrix, to a much smaller problem expressed in terms of a few scalar variables representing intra‑core, core‑to‑petal, and intra‑petal edge weights. The SDP constraints—row‑stochasticity, symmetry, and positive semidefiniteness of (W - \frac{1}{N}{\bf 1}{\bf 1}^T)—are then written in block‑diagonal form.

A key technical contribution is the inductive comparison of characteristic polynomials that arise from the Karush‑Kuhn‑Tucker (KKT) slackness conditions. By examining the polynomial equations for successive strata, the authors are able to solve for the coefficients recursively, ultimately obtaining explicit formulas for the optimal weights. For the symmetric petal network the optimal weight on edges linking the central node to a petal is (\frac{2}{d+2}) (where (d) is the degree of each petal), and all intra‑petal edges receive the same weight. In the CCS case the core‑core edges receive weight (\frac{1}{k}) (with (k) core nodes) while core‑to‑petal edges receive (\frac{2}{k+d+2}). These expressions are linear in the structural parameters, making them easy to compute for any network size.

The paper also introduces a class of “special leaves” that consist of multiple sub‑petals sharing a common attachment node. Such configurations cannot be handled by the basic stratification alone; the authors treat each special leaf as a super‑node and derive its optimal weight separately. This extension demonstrates the flexibility of the framework and shows that the analytical solution remains valid even when the leaf structures become more intricate.

From a spectral perspective, the optimal weight matrix maximizes the second smallest eigenvalue (algebraic connectivity) of the Laplacian, which directly controls the convergence rate of the averaging process. The authors prove that their closed‑form weights achieve this maximal eigenvalue, thereby meeting the theoretical lower bound on convergence time for the given topology.

Numerical experiments corroborate the theory. Simulations on networks ranging from a few hundred to several thousand nodes show that the analytically derived weights reduce the number of iterations needed to reach a given error tolerance by 30–50 % compared with uniform weighting or with weights obtained from generic SDP solvers. Moreover, the computational effort to evaluate the closed‑form formulas scales linearly with the number of nodes, confirming that the method is suitable for real‑time implementation in large‑scale sensor deployments.

In conclusion, the authors provide a complete analytical solution to the FDC problem for two important families of petal networks. By marrying stratification with SDP and exploiting the symmetry‑induced polynomial structure of the KKT conditions, they obtain explicit optimal weights, extend the approach to complex leaf configurations, and validate the results both theoretically and experimentally. The framework is promising for extension to other hierarchical or clustered graph topologies, and it opens avenues for future work that incorporates asynchronous updates, time‑varying links, or communication delays while preserving the analytical tractability demonstrated here.