---
title: Stochastic Externally Equitable Partitions
url: https://www.emergentmind.com/topics/stochastic-externally-equitable-partitions-seeps
type: topic
---

# Stochastic Externally Equitable Partitions

A stochastic externally equitable partition (sEEP) is a block-structured node partition in a graph or network ensemble such that, on average, every node within a block exhibits the same expected pattern of external connections to each other block. This property generalizes externally equitable partitions from deterministic graphs to settings where the adjacency matrix itself is a random variable, capturing role structure as statistical regularity in connectivity profiles rather than as strict combinatorial constraints. sEEPs are central to principled definitions of node roles, the analysis of block-regular graph ensembles, and the design of algorithms for robust role discovery in the presence of noise and stochasticity [2305.19087, 1610.02668].

## 1. Fundamental Definitions

In a graph $G=(V,E)$ with adjacency matrix $A \in \mathbb{R}^{n \times n}$, a partition $\mathcal{C} = \{C_1, ..., C_k\}$ is externally equitable if, for all $v,u \in C_i$ and all $j \ne i$, the number of neighbors in $C_j$ is invariant:
$$
\sum_{x \in N(v) \cap C_j} 1 = \sum_{x \in N(u) \cap C_j} 1\,, \quad \forall j \ne i \,.
$$
Algebraically, with indicator matrix $H$ for $\mathcal{C}$, and with $A^{(e)}$ obtained from $A$ by zeroing out intra-block edges,
$$
A^{(e)} H = H A^{(\pi,e)}
$$
for some $k \times k$ matrix $A^{(\pi,e)}$.

A stochastic externally equitable partition (sEEP) emerges when $A$ is random, e.g., drawn from a distribution over graphs. Let $\mu = \mathbb{E}[A]$; then $\mathcal{C}$ is an sEEP if:
$$
\mu^{(e)} H = H \Theta
$$
for some $k \times k$ matrix $\Theta$. Equivalently, nodes in the same block have the same expected external connectivity to each other block:
$$
\mathbb{E}\left[ \sum_{x \in N(v) \cap C_j} A_{v,x} \right] = \mathbb{E}\left[ \sum_{x \in N(u) \cap C_j} A_{u,x} \right], \quad v, u \in C_i,\, j \ne i\,.
$$
In the equitable random graph ensemble $G(N, \{N_a\}, B)$, these expectations are enforced exactly for every node and block [1610.02668].

## 2. Equitable Ensembles and Block-Regular Structures

The equitable graph ensemble consists of undirected graphs on $N$ vertices, partitioned into $m$ blocks $\mathbf{B} = \{B_1, ..., B_m\}$ with $N_a = |B_a|$. The block-regularity matrix $B = (b_{ij})$ imposes that every vertex in block $B_i$ has exactly $b_{ij}$ neighbors in block $B_j$. Degree-regularity requires $N_r b_{rs} = N_s b_{sr}$, and the set of all graphs satisfying these constraints forms the uniform (microcanonical) ensemble $\mathcal{G}(N, \{N_a\}, B)$.

An sEEP is intrinsic to these ensembles by construction: the partition into $\mathbf{B}$ yields blocks whose external degrees are identically distributed, so the mean adjacency matrix $\mathbb{E}[A]$ exhibits the externally equitable structure, and every sampled realization strictly matches the sEEP pattern [1610.02668].

## 3. Optimization Frameworks for Approximating sEEPs

In practical network data, neither EPs nor EEPs are typically exact, motivating optimization-based role recovery. Two key cost functions are introduced [2305.19087]:

- **One-step cost:** 
  $$
  \Gamma_{EP}(A, H) = \| A H - H (H^{\top}H)^{-1} H^{\top}A H \|
  $$
  If $\Gamma_{EP}(A, H) = 0$, $H$ defines an EP. The EEP variant restricts attention to external (off-diagonal) blocks.

- **d-step (deep) cost:** 
  $$
  \Gamma_{d-EP}(A, H) = \sum_{t=1}^d \rho(A)^{-t} \| A^t H - H (H^{\top}H)^{-1} H^{\top}A^t H \|
  $$
  where $\rho(A)$ is the spectral radius. This measures deviation from equitability over walks of length up to $d$.

For $d \to \infty$, with an $\ell_1$ norm, the minimizer clusters the dominant eigenvector into $k$ groups via 1D $k$-means [2305.19087]. For the short-term cost, a practical (NP-hard) heuristic alternates between “continuous” Weisfeiler–Leman refinement and block re-clustering (e.g., via $k$-means or average linkage).

## 4. Theoretical Guarantees for sEEP Recovery

Long-term (deep) cost minimization possesses an explicit optimum: when the adjacency matrix $A$ has a unique top eigenvector, clustering its entries yields the optimal partition for $\lim_{d \to \infty}\Gamma_{d-EP}(A,H)$, with computational complexity $O(a + n k + n \log n)$ where $a$ is the cost of eigenvector computation [2305.19087].

Under a planted role model—such as the role-infused partition (RIP) stochastic block model—provided blocks are sufficiently large that binomial fluctuations do not overwhelm the separation between role profiles, the continuous Weisfeiler–Leman-inspired algorithm provably recovers planted roles with high probability:
$$
n > -\frac{9\,W_{-1}\left((q-1)\,\delta^2 / (9 k^2)\right)}{2 \delta^2}
$$
where $\delta$ is the minimum $\ell_2$-distance between any two role profiles and $W_{-1}$ is the lower real branch of the Lambert $W$ function [2305.19087].

## 5. Spectral Properties and Detectability in Equitable Graphs

For equitable ensembles with two equal-sized blocks ($m=2$), the block structure is encoded in a $2 \times 2$ quotient matrix:
$$
B = \begin{pmatrix}
c_{\text{in}} & c_{\text{out}} \\
c_{\text{out}} & c_{\text{in}}
\end{pmatrix}
$$
The “community” eigenvalue $\lambda_{\text{com}} = c_{\text{in}} - c_{\text{out}}$ determines the separability of block structure, while the spectral density of $A$ exhibits a Kesten–McKay bulk:
$$
\rho(\lambda) = \frac{c \sqrt{4(c-1)-\lambda^2}}{2\pi (c^2 - \lambda^2)}, \quad |\lambda| \leq 2\sqrt{c-1}
$$
where $c = c_{\text{in}} + c_{\text{out}}$ [1610.02668].

Block recovery is efficient:
- If $|\lambda_{\text{com}}| > 2\sqrt{c-1}$, the informative eigenvalue is separated from the bulk, and a sign-clustering of its eigenvector recovers the sEEP exactly.
- In contrast to ordinary stochastic block models, the equitable ensemble lacks a sharp detectability threshold; the planted partition is always recoverable efficiently by spectral means for any $\lambda_{\text{com}}\neq0$ due to rigid block-regularity [1610.02668].

## 6. Role-Infused Partition Benchmark and Validation

The role-infused partition (RIP) model is designed for benchmarking sEEP detection [2305.19087]. Nodes are labeled by community $r$, role $\ell$, and copy $s$, and the adjacency matrix is defined by parameters: community count $c$, roles $k$, nodes per role-community $n$, background probability $p$, and role interaction matrix $\Omega_{\text{role}}$. Edges internal to a community are sampled according to $\Omega_{\text{role}}$, while inter-community links are assigned with probability $p$.

The expectation of the adjacency matrix satisfies:
$$
\mathbb{E}[A] = H_B \Omega H_B^{\top}
$$
where $H_B$ is the indicator matrix for the $ck$ blocks and $\Omega$ defines the expected connectivity. The planted partition into $k$ roles is the coarsest externally equitable partition (an sEEP) of $\mathbb{E}[A]$.

Optimization-based algorithms reliably recover roles in the RIP model, interpolating between community-based and role-based partitions, validating the concept and practical utility of sEEPs [2305.19087].

## 7. Implications and Applications

Minimizing the aforementioned cost functions on observed adjacency matrices, or their sample means in multilayer or dynamic settings, yields approximate solutions to the sEEP-recovery problem. Both eigenvector-based and Weisfeiler–Leman-like heuristics succeed on block-regular and role-infused benchmark networks. Empirically, these methods recover latent roles under noise, supporting the use of sEEPs for robust role extraction in real-world networks. Discovered roles serve as inputs for downstream tasks such as graph-level embeddings, few-shot learning, and dynamical systems reduction, indicating that most-equitable partitions encode meaningful structural and functional information [2305.19087].

In summary, sEEPs provide a mathematically rigorous and computationally tractable framework for defining, detecting, and validating node role structure robustly, bridging the gap between exact graph symmetries and the statistical regularities of noisy, real-world networks [2305.19087, 1610.02668].

Source: https://www.emergentmind.com/topics/stochastic-externally-equitable-partitions-seeps