Cluster size distributions of discrete random fields
Dan Cheng, John Ginos · 2026
Read Cluster size distributions of discrete random fields by Dan Cheng, John Ginos (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-14586.txt.
Introduction
Dan Cheng and John Ginos study random fields on a grid in this 2026 paper. A discrete random field assigns a random value to each point of the d-dimensional integer lattice. Fix a threshold u. The excursion set, all points where the value exceeds u, decomposes into connected components, or clusters. Each cluster's size is the number of lattice points it contains, and that size is random.
The paper investigates the probability distribution of these cluster sizes. Cluster size laws matter wherever high spots group together, from brain imaging to materials. Readers get theory for how often big clusters appear by chance in a lattice field.
Catalog record
| CREATOR | Dan Cheng, John Ginos |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 66 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2601-14586
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Cluster size distributions of discrete random fields Dan Cheng, John Ginos Abstract We study discrete random fields { X t : t ∈ ℤ d } \{X_{t}:t\in{\mathbb{Z}}^{d}\} parameterized on the d d -dimensional integer lattice ℤ d {\mathbb{Z}}^{d} . For a fixed threshold u u , the excursion set { t ∈ ℤ d : X t > u } \{t\in{\mathbb{Z}}^{d}:X_{t}>u\} decomposes into connected components or clusters, whose size, defined as the number of lattice points they contain, are random. This paper investigates the probability distribution of these cluster sizes. For stationary random fields, we derive exact expressions for the cluster size distribution. To address nonstationary settings, we introduce a peak-based cluster size distribution, which characterizes the distribution of cluster sizes conditional on the presence of a local maximum above u u . This formulation provides a tractable alternative when exact cluster size distributions are analytically inaccessible. The proposed framework applies broadly to Gaussian and non-Gaussian random fields, relying only on their joint dependence structure. Our results provide a theoretical foundation for quantifying spatial extent in discretely sampled data, with applications to medical imaging, geoscience, environmental monitoring, and other scientific areas where thresholded random fields naturally arise. Keywords : random fields; discrete lattice; spatial extent; connected components; cluster size; peak-based; stationary; nonstationary; threshold. Mathematics Subject Classification : 62M40, 62G32, 62E15, 60G70, 60G60. 1 Introduction Random fields play a central role in modern spatial statistics, stochastic modeling, and signal detection. They provide a natural framework for characterizing spatially correlated data across a wide range of scientific disciplines, including medical imaging, geoscience, environmental studies, and cosmology [ 2 , 5 , 9 , 16 , 19 , 24 ] . When a random field is observed over a discrete lattice such as {\mathbb{Z}}^{d} , the resulting process is referred to as a discrete random field . A fundamental problem in the analysis of such fields concerns the geometric and topological properties of regions where the field exceeds a fixed threshold level u . These regions, commonly called excursion sets, decompose into random connected components, or clusters, formed by neighboring lattice sites above the threshold. Understanding the statistical distribution of cluster sizes in random fields is fundamental for both theory and applications. In spatial statistics and random field theory, the cluster size distribution under a null (pure noise) model provides a benchmark for assessing the significance of observed spatial patterns. For example, in neuroimaging, random field theory underlies cluster-based inference procedures for detecting activation regions in fMRI data [ 3 , 10 , 14 , 16 , 20 , 23 , 25 ] . Similar principles arise in geoscience and cosmology, where cluster size distributions describe the spatial extent of anomalies in correlated data [ 1 , 4 , 9 , 18 , 21 ] . Recent work in climate and environmental statistics has emphasized the importance of spatial extent inference, which aims to quantify the areal footprint of extreme events [ 22 , 26 ] . In parallel, inference for threshold exceedances in random fields has been applied to spatiotemporal extreme value analysis [ 15 , 17 ] . While the geometry of excursion sets for continuous Gaussian fields has been extensively studied through tools such as the Gaussian kinematic formula and Euler characteristic heuristics [ 2 , 24 ] , exact cluster-size distributions are typically intractable in continuous settings. In contrast, for discrete random fields on lattices, it is possible to formulate and compute cluster size distributions explicitly under suitable assumptions. The discrete setup allows a combinatorial characterization of connected components, leading to exact expressions in low dimensions and numerically tractable formulations in higher dimensions. In this paper, we investigate the cluster size distribution of discrete stationary random fields \{X_{t}:t\in{\mathbb{Z}}^{d}\} defined on the d -dimensional integer lattice {\mathbb{Z}}^{d} . We focus on the probabilistic structure of clusters that arise arising from thresholding the field at a fixed level u\in{\mathbb{R}} . Clusters are defined as connected components of the excursion set \{t\in{\mathbb{Z}}^{d}:X_{t}>u\} . Each cluster above the threshold u , denoted by C_{u} , has a random size |C_{u}| , defined simply as the number of lattice points in the component. Illustrations in one and two dimensions are provided in Figures 1 and 3 . Our first objective is to derive the exact cluster size distribution \begin{split}{\mathbb{P}}(|C_{u}|=k\ |\ \text{a cluster $C_{u}$ above $u$ exists}),\quad k=1,2,\ldots,\end{split} for stationary random fields, where the covariance structure depends only on spatial lag. Motivated by recent advances in peak inference for random fields [ 6 , 7 , 8 ] and peaks-over-threshold methods [ 15 , 17 ] , we further introduce and develop the peak-based cluster size distribution . This characterizes the distribution of cluster sizes conditional on the presence of a local maximum above u . Specifically, we derive the conditional probability {\mathbb{P}}(|C_{u}(t)|=k\ |\ t\ \text{\rm is a local maximum and }X_{t}>u),\quad k=1,2,\ldots, where C_{u}(t) denotes the cluster containing an observed local maximum t that exceeds the threshold u . This formulation connects local peak statistics with cluster extent, bridging ideas from spatial extreme value theory and classical random field analysis. It is particularly valuable in nonstationary settings, where local variations in the covariance structure influence both the shape and spatial extent of clusters. In contrast, obtaining exact cluster size distributions in the nonstationary case appears analytically intractable. In many applications, data are observed on discrete grids rather than as continuous fields. The framework developed in this paper thus provides a unified theoretical foundation for cluster modeling and inference in discrete random fields. Beyond its intrinsic theoretical interest, the results have broad practical implications for signal detection and spatial extent inference in correlated lattice data. Applications include neuroimaging [ 3 , 10 , 14 , 16 , 23 ] , geophysical anomaly detection [ 9 , 21 ] , and climate data analysis [ 11 , 13 ] , where discretely sampled random fields are ubiquitous. Finally, while Gaussian random fields serve as a convenient and widely used reference model, the general framework developed here does not require Gaussianity. As will become clear from the arguments below, the derivations depend only on the joint dependence structure of the field. Consequently, the methodology applies to broad classes of non-Gaussian random fields, provided their finite-dimensional distributions are well defined. 2 The exact cluster size distribution We begin with stationary discrete random fields and derive exact expressions for their cluster size distributions. In general, the analysis of random fields indexed by {\mathbb{Z}}^{d} is conceptually rich and technically demanding. To present the main ideas in a transparent manner, we first focus on the one-dimensional case, where the structure of clusters and the required joint probabilities admit explicit characterization. Simulated Gaussian process and clusters above u=0.5 Empirical cluster size distribution Figure 1: The left panel is the scatter plot of a simulated realization of a centered stationary Gaussian process with covariance C(t,s)=e^{-(t-s)^{2}} , where exceedances above the threshold u=0.5 are marked in red, and the corresponding clusters are marked in consecutive red dots along the index {\mathbb{Z}} . This realization exhibits 14 observed clusters: 7 of size one, 5 of size two, 1 of size three, and 1 of size four. The right panel displays the empirical cluster size distribution