arxiv: 2605.10225 · v1 · submitted 2026-05-11 · 🧮 math.ST · stat.ME· stat.TH

Recognition: 1 theorem link

· Lean Theorem

Increasing domain asymptotics for covariate-based nonparametric Bayesian intensity estimation with Gaussian and Besov-Laplace priors

Matteo Giordano, Patric Dolmeta

Pith reviewed 2026-05-12 03:22 UTC · model grok-4.3

classification 🧮 math.ST stat.MEstat.TH

keywords nonparametric Bayesian estimationpoint process intensityGaussian priorsBesov-Laplace priorsincreasing domain asymptoticsposterior contraction ratesergodic covariatesspatial inhomogeneity

0 comments

The pith

Gaussian priors with link functions achieve minimax-optimal posterior contraction rates for estimating intensities of covariate-driven point processes over expanding domains.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper studies Bayesian nonparametric estimation of the intensity function for point processes that depend on observed covariates, with data collected over increasingly large spatial windows. It shows that a broad family of Gaussian priors, used together with flexible link functions, attains the fastest possible rates at which the posterior distribution concentrates around the true intensity when the covariates satisfy an ergodicity condition. The authors further establish that Besov-Laplace priors deliver optimal rates for intensities that vary spatially and lie in Besov spaces of low integrability. These results rest on a general posterior concentration theorem that extends earlier work. Simulations and two real-data examples from forestry and environmental science illustrate the procedures.

Core claim

Under increasing domain asymptotics and ergodicity of the covariates, a wide class of Gaussian priors combined with flexible link functions achieves minimax-optimal posterior contraction rates for the intensity function of a covariate-driven point process. Besov-Laplace priors further yield optimal estimation of spatially inhomogeneous intensities belonging to Besov spaces with low integrability index. Both sets of results follow from a general concentration theorem that extends recent findings in the literature.

What carries the argument

General concentration theorem for posterior distributions of the intensity, applied to Gaussian priors and Besov-Laplace priors together with link functions.

Load-bearing premise

The covariates are assumed to be ergodic so that their spatial averages converge to population expectations over large domains.

What would settle it

A concrete ergodic covariate sequence for which the posterior contraction rate falls strictly below the minimax rate, or a non-ergodic sequence where the claimed optimality still holds, would settle whether the rates are truly optimal under the stated conditions.

Figures

Figures reproduced from arXiv: 2605.10225 by Matteo Giordano, Patric Dolmeta.

**Figure 2.** Figure 2: Left to right: Posterior means for Gaussian (solid green) and Besov-Laplace [PITH_FULL_IMAGE:figures/full_fig_p013_2.png] view at source ↗

**Figure 3.** Figure 3: Left to right: Posterior means for Gaussian (solid green) and Laplace (solid [PITH_FULL_IMAGE:figures/full_fig_p016_3.png] view at source ↗

**Figure 4.** Figure 4: Terrain elevation (in meters) and steepness (norm of the altitude gradient) in [PITH_FULL_IMAGE:figures/full_fig_p019_4.png] view at source ↗

**Figure 5.** Figure 5: Estimates of the intensity as a function of terrain elevation and slope, respec [PITH_FULL_IMAGE:figures/full_fig_p020_5.png] view at source ↗

**Figure 6.** Figure 6: Posterior mean for the Besov-Laplace priors (with regularity [PITH_FULL_IMAGE:figures/full_fig_p021_6.png] view at source ↗

**Figure 7.** Figure 7: Plug-in posterior mean for the Besov-Laplace priors (with regularity [PITH_FULL_IMAGE:figures/full_fig_p021_7.png] view at source ↗

read the original abstract

We study the problem of estimating the intensity function of a covariate-driven point process based on observations of the points and covariates over a large window. We consider the nonparametric Bayesian approach, and show that a wide class of Gaussian priors, combined with flexible link functions, achieves minimax-optimal posterior contraction rates in the increasing domain asymptotics and under the assumption that the covariates be ergodic. We also employ Besov-Laplace priors, which are popular in imaging and inverse problems due to their edge-preserving and sparsity-promoting properties. We prove that these yield optimal estimation of spatially inhomogeneous intensities belonging to Besov spaces with low integrability index. These results are based on a general concentration theorem that extends recent findings from the literature. To corroborate the theory, we provide extensive numerical simulations, implementing the considered procedures via suitable posterior sampling schemes. Further, we present two real data analyses motivated by applications in forestry and the environmental sciences.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper establishes minimax-optimal posterior contraction rates for Bayesian nonparametric intensity estimation in covariate-driven point processes under increasing domains, using Gaussian priors with links and Besov-Laplace priors, assuming ergodic covariates.

read the letter

This work gives posterior contraction rates for estimating the intensity of a point process driven by covariates, observed over an expanding window. Gaussian priors with flexible links and Besov-Laplace priors both reach the minimax rate in this increasing-domain regime when the covariates are ergodic. The Besov-Laplace part targets intensities in Besov spaces with low integrability, which fits cases with spatial inhomogeneity or edges. That extension from earlier concentration theorems is the main new piece, and the paper backs it with simulations plus two real-data examples in forestry and environmental science. Those practical sections are a plus; they show how to sample from the posterior and apply the methods to actual spatial data. The ergodicity assumption carries a lot of weight for controlling the empirical covariate measure and any link-function bias. The abstract does not spell out a quantitative mixing rate or explicit remainder term in the ergodic theorem, so it is not clear whether the general concentration result absorbs the approximation error at the exact nonparametric speed. If the ergodicity is only weak, the claimed rates could slip. The proofs rest on extending recent literature rather than starting from scratch, and the citation pattern looks standard for this corner of Bayesian nonparametrics and spatial statistics. No obvious circularity or invented quantities appear. This is useful for people working on Bayesian methods for spatial point processes or environmental applications who already know the contraction-rate literature. It is worth sending to peer review so the details of the ergodicity handling and the simulation setups can be checked directly.

Referee Report

2 major / 2 minor

Summary. The manuscript develops nonparametric Bayesian methods for estimating the intensity function of a covariate-driven point process observed over a large domain. It establishes minimax-optimal posterior contraction rates for a class of Gaussian priors with flexible link functions under ergodicity assumptions on the covariates in the increasing domain regime. Additionally, it shows that Besov-Laplace priors achieve optimal rates for intensities in Besov spaces with low integrability indices. The theoretical results rely on a general concentration theorem extension, supported by simulations and real data analyses in forestry and environmental sciences.

Significance. If the results hold, this provides important theoretical foundations for Bayesian nonparametric inference in spatial point processes with covariates, extending concentration theorems to the increasing-domain setting with ergodic covariates. Credit is due for the general concentration theorem approach, the inclusion of Besov-Laplace priors for their edge-preserving properties, and the provision of posterior sampling schemes with simulations and real-data examples. The significance is tempered by the need for precise quantitative control under the ergodicity assumption.

major comments (2)

[Abstract and main theorems] Abstract and the statement of the main results (likely §4): The claim of minimax-optimal posterior contraction for Gaussian priors under ergodic covariates rests on extending a general concentration theorem, but the abstract and theorem statements provide no indication of a quantitative ergodic theorem with explicit remainder terms. Without this, the bias term from the flexible link function cannot be guaranteed to be controlled at the nonparametric rate.
[Section 3] The extension of the concentration theorem (Section 3): The transfer of rates to the covariate-driven intensity estimation requires the empirical covariate measure to converge sufficiently fast; if only qualitative ergodicity is assumed without mixing rates matching the contraction rate ε_n, the optimality claim is at risk. A concrete test would involve deriving the remainder under specific α-mixing conditions.

minor comments (2)

[Abstract] The description of the 'wide class of Gaussian priors' in the abstract could specify the covariance kernels or hyperparameter ranges more explicitly to clarify the scope.
[Numerical simulations] The simulations section would benefit from additional details on the covariate process generation and the exact posterior sampling implementation to enhance reproducibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading and constructive comments, which help clarify the presentation of the ergodicity assumptions. We address each major comment below and will incorporate clarifications in the revised manuscript.

read point-by-point responses

Referee: [Abstract and main theorems] Abstract and the statement of the main results (likely §4): The claim of minimax-optimal posterior contraction for Gaussian priors under ergodic covariates rests on extending a general concentration theorem, but the abstract and theorem statements provide no indication of a quantitative ergodic theorem with explicit remainder terms. Without this, the bias term from the flexible link function cannot be guaranteed to be controlled at the nonparametric rate.

Authors: We agree that explicit mention of quantitative control would strengthen the statements. The general concentration theorem (Section 3) is formulated so that the deviation of the empirical covariate measure from its expectation is absorbed into the remainder term, which is controlled at rate o(ε_n) under the ergodicity assumption via standard results on spatial averages (e.g., the ergodic theorem for stationary random fields). The bias induced by the link function is then dominated by the nonparametric rate. To address the concern directly, we will revise the abstract and the statements of Theorems 4.1–4.3 to include a short parenthetical note referencing the quantitative ergodic convergence that ensures the remainder is negligible relative to ε_n. revision: partial
Referee: [Section 3] The extension of the concentration theorem (Section 3): The transfer of rates to the covariate-driven intensity estimation requires the empirical covariate measure to converge sufficiently fast; if only qualitative ergodicity is assumed without mixing rates matching the contraction rate ε_n, the optimality claim is at risk. A concrete test would involve deriving the remainder under specific α-mixing conditions.

Authors: This observation is well taken. The manuscript currently invokes ergodicity in the qualitative sense that the empirical measure converges almost surely to the invariant measure, which is standard for increasing-domain asymptotics. To provide the requested concrete test, we will add a remark (or short appendix subsection) deriving an explicit bound on the remainder under α-mixing conditions on the covariate field with mixing coefficients decaying at a polynomial rate sufficient to be o(ε_n). This will confirm that the optimality claim holds under mixing rates compatible with typical nonparametric contraction rates. revision: yes

Circularity Check

0 steps flagged

No significant circularity; central results rest on extension of general concentration theorem plus ergodicity assumption

full rationale

The paper derives posterior contraction rates from a general concentration theorem that extends recent literature, under the ergodicity assumption on covariates in the increasing-domain regime. No quoted equations or steps reduce the claimed minimax rates for Gaussian or Besov-Laplace priors to fitted parameters, self-definitions, or load-bearing self-citations by construction. The derivation chain remains self-contained against external benchmarks once the stated assumptions hold.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

Central claims rest on ergodicity of covariates as a domain assumption and extension of a general concentration theorem from recent literature; no free parameters or invented entities specified.

axioms (2)

domain assumption Covariates are ergodic
Required for increasing domain asymptotics to yield the stated contraction rates.
standard math General concentration theorem extends recent findings from the literature
Basis for proving the posterior contraction results.

pith-pipeline@v0.9.0 · 5459 in / 1148 out tokens · 56029 ms · 2026-05-12T03:22:13.749683+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear
posterior contraction rates ... Gaussian priors ... Besov-Laplace priors ... general concentration theorem ... ergodic covariates

Reference graph

Works this paper leans on

89 extracted references · 89 canonical work pages

[1]

ANNALS OF STATISTICS , volume=

Asymptotic behaviour of the empirical bayes posteriors associated to maximum marginal likelihood estimator , author=. ANNALS OF STATISTICS , volume=

work page
[2]

Adams, Ryan Prescott and Murray, Iain and MacKay, David J. C. , booktitle =. Tractable nonparametric

work page
[3]

Rates of contraction of posterior distributions based on p -exponential priors , author=

work page
[4]

Adaptive inference over

Agapiou, Sergios and Savva, Aimilia , journal=. Adaptive inference over

work page
[5]

Agapiou, Sergios and Wang, Sven , journal=

work page
[6]

Lember and A

J. Lember and A. van der Vaart , journal =. On universal. 2007 , doi =

work page 2007
[7]

Electron

van Waaij, Jan and van Zanten, Harry , doi =. Electron. J. Stat. , number =

work page
[8]

Nonparametric estimation of the dependence of a spatial point process on spatial covariates , volume =

Baddeley, Adrian and Chang, Ya-Mei and Song, Yong and Turner, Rolf , year =. Nonparametric estimation of the dependence of a spatial point process on spatial covariates , volume =

work page
[9]

2016 , publisher=

Spatial point patterns: methodology and applications with R , author=. 2016 , publisher=

work page 2016
[10]

Bandyopadhyay, Soutir and Subba Rao, Suhasini , TITLE =. J. Roy. Statist. Soc. Ser. B , FJOURNAL =. 2017 , NUMBER =

work page 2017
[11]

Probability theory and related fields , volume=

Risk bounds for model selection via penalization , author=. Probability theory and related fields , volume=. 1999 , publisher=

work page 1999
[12]

Belitser, Eduard and Serra, Paulo and van Zanten, Harry , TITLE =. J. Statist. Plann. Inference , FJOURNAL =. 2015 , PAGES =

work page 2015
[13]

Anisotropic function estimation using multi-bandwidth

Bhattacharya, Anirban and Pati, Debdeep and Dunson, David , journal=. Anisotropic function estimation using multi-bandwidth

work page
[14]

2004 , doi =

Lucien Birg. 2004 , doi =

work page 2004
[15]

Borrajo, M. I. and Gonz\'. Bootstrapping kernel intensity estimation for inhomogeneous point processes with spatial covariates , volume =. Comput. Statist. Data Anal. , pages =

work page
[16]

, booktitle =

Brillinger, David R. , booktitle =. Comparative aspects of the study of ordinary time series and of point processes , year =

work page
[17]

Isma. Ann. Statist. , number =

work page
[18]

Statistical Science , author=

work page
[19]

Cox, D. R. , TITLE =. J. Roy. Statist. Soc. Ser. B , FJOURNAL =. 1955 , PAGES =

work page 1955
[20]

Cressie, Noel A. C. , TITLE =. 2015 , PAGES =

work page 2015
[21]

2003 , publisher=

An introduction to the theory of point processes: volume I: elementary theory and methods , author=. 2003 , publisher=

work page 2003
[22]

Dashti, Masoumeh and Harris, Stephen and Stuart, Andrew , journal=

work page
[23]

and Kass, Robert E

DiMatteo, Ilaria and Genovese, Christopher R. and Kass, Robert E. , TITLE =. 2001 , NUMBER =

work page 2001
[24]

On the Consistency of

Diaconis, Persi and Freedman, David , journal=. On the Consistency of. 1986 , publisher=

work page 1986
[25]

Diggle , issn =

Peter J. Diggle , issn =. A point process modelling approach to raised incidence of a rare phenomenon in the vicinity of a prespecified point , urldate =. J. Roy. Statist. Soc. Ser. A , number =. 1990 , bdsk-url-1 =

work page 1990
[26]

, edition =

Diggle, Peter J. , edition =. Statistical Analysis of Spatial and Spatio-Temporal Point Patterns , volume =

work page
[27]

Diggle and Nicholas Lange and Francine M

Peter J. Diggle and Nicholas Lange and Francine M. Beneš , journal =. Analysis of Variance for Replicated Spatial Point Patterns in Clinical Neuroanatomy , urldate =

work page
[28]

A nonparametric

Dolmeta, Patric and Giordano, Matteo , journal=. A nonparametric

work page
[29]

2025 , organization=

Dolmeta, Patric and Giordano, Matteo , booktitle=. 2025 , organization=

work page 2025
[30]

Supplement to ``Increasing domain asymptotics for covariate-based nonparametric

Dolmeta, Patric and Giordano, Matteo , journal=. Supplement to ``Increasing domain asymptotics for covariate-based nonparametric

work page
[31]

1994 , publisher=

Ideal spatial adaptation by wavelet shrinkage , author=. 1994 , publisher=

work page 1994
[32]

1998 , publisher=

Minimax estimation via wavelet shrinkage , author=. 1998 , publisher=

work page 1998
[33]

Posterior concentration rates for counting processes with

Donnet, Sophie and Rivoirard, Vincent and Rousseau, Judith and Scricciolo, Catia , doi =. Posterior concentration rates for counting processes with. 2017 , bdsk-url-1 =

work page 2017
[34]

Annales Henri Lebesgue , volume=

Multiscale functional inequalities in probability: Constructive approach , author=. Annales Henri Lebesgue , volume=

work page
[35]

2010 , publisher=

Partial Differential Equations , author=. 2010 , publisher=

work page 2010
[36]

and van der Vaart, Aad W

Ghosal, Subhashis and Ghosh, Jayanta K. and van der Vaart, Aad W. , date-added =. Convergence rates of posterior distributions , volume =. Ann. Statist. , number =

work page
[37]

Fundamentals of nonparametric

Ghosal, Subhashis and Van der Vaart, Aad W , year=. Fundamentals of nonparametric

work page
[38]

Semiparametric

Giordano, Matteo and Ray, Kolyan , journal=. Semiparametric

work page
[39]

Mathematical foundations of infinite-dimensional statistical models , year =

Gin\'e, Evarist and Nickl, Richard , date-added =. Mathematical foundations of infinite-dimensional statistical models , year =

work page
[40]

2023 , publisher=

Giordano, Matteo , journal=. 2023 , publisher=

work page 2023
[41]

Bernoulli , volume=

Nonparametric Bayesian intensity estimation for covariate-driven inhomogeneous point processes , author=. Bernoulli , volume=. 2026 , publisher=

work page 2026
[42]

Consistency of

Giordano, Matteo and Nickl, Richard , journal=. Consistency of. 2020 , publisher=

work page 2020
[43]

Nonparametric

Giordano, Matteo and Ray, Kolyan , journal=. Nonparametric. 2022 , publisher=

work page 2022
[44]

Posterior contraction rates for

Grant, James A and Leslie, David S , journal=. Posterior contraction rates for

work page
[45]

On consistent nonparametric intensity estimation for inhomogeneous spatial point processes , volume =

Guan, Yongtao , fjournal =. On consistent nonparametric intensity estimation for inhomogeneous spatial point processes , volume =. J. Amer. Statist. Assoc. , mrclass =

work page
[46]

A note on non-parametric

Gugushvili, Shota and Spreij, Peter , journal=. A note on non-parametric

work page
[47]

Fast and scalable non-parametric

Gugushvili, Shota and Van Der Meulen, Frank and Schauer, Moritz and Spreij, Peter , journal=. Fast and scalable non-parametric

work page
[48]

Maximum a posteriori probability estimates in infinite-dimensional

Helin, Tapio and Burger, Martin , journal=. Maximum a posteriori probability estimates in infinite-dimensional. 2015 , publisher=

work page 2015
[49]

An Adaptive

Haario, Heikki and Saksman, Eero and Tamminen, Johanna , year =. An Adaptive

work page
[50]

2012 , publisher=

Wavelets, approximation, and statistical applications , author=. 2012 , publisher=

work page 2012
[51]

2002 , publisher=

Random rates in anisotropic regression (with a discussion and a rejoinder by the authors) , author=. 2002 , publisher=

work page 2002
[52]

Statistical analysis and modelling of spatial point patterns , year =

Illian, Janine and Penttinen, Antti and Stoyan, Helga and Stoyan, Dietrich , isbn =. Statistical analysis and modelling of spatial point patterns , year =

work page
[53]

Illian, Janine B. and S. A toolbox for fitting complex spatial point process models using integrated nested. Ann. Appl. Stat. , mrclass =. 2012 , bdsk-url-1 =. doi:10.1214/11-AOAS530 , fjournal =

work page doi:10.1214/11-aoas530 2012
[54]

Asymptotic normality of estimates in spatial point processes , author=. Scand. J. Statist. , pages=. 1993 , volume=

work page 1993
[55]

Point processes and queuing problems (Colloqium, Keszthely, 1978) , volume=

Central limit theorem and convergence of empirical processes for stationary point processes , author=. Point processes and queuing problems (Colloqium, Keszthely, 1978) , volume=. 1981 , publisher=

work page 1978
[56]

Stochastic Environmental Research and Risk Assessment , volume=

Pinpointing spatio-temporal interactions in wildfire patterns , author=. Stochastic Environmental Research and Risk Assessment , volume=

work page
[57]

, TITLE =

Karr, Alan F. , TITLE =. 1991 , PAGES =

work page 1991
[58]

Optimality of

Kirichenko, Alisa and van Zanten, Harry , fjournal =. Optimality of. J. Mach. Learn. Res. , mrclass =

work page
[59]

Sparsity-promoting

Kolehmainen, Ville and Lassas, Matti and Niinim. Sparsity-promoting. Inverse Probl. , volume=. 2012 , publisher=

work page 2012
[60]

The Annals of Applied Statistics , number =

Jonathan Koh and Fran. The Annals of Applied Statistics , number =

work page
[61]

Kottas, Athanasios and Sans\'. J. Statist. Plann. Inference , FJOURNAL =. 2007 , NUMBER =

work page 2007
[62]

Kuo and S

L. Kuo and S. K. Ghosh , date-added =

work page
[63]

Kutoyants, Yu. A. , isbn =. Statistical inference for spatial. 1998 , bdsk-url-1 =

work page 1998
[64]

Discretization invariant

Lassas, M and Saksman, E and Siltanen, S , journal=. Discretization invariant. 2009 , publisher=

work page 2009
[65]

Approximation, metric entropy and small ball estimates for

Li, Wenbo V and Linde, Werner , journal=. Approximation, metric entropy and small ball estimates for. 1999 , publisher=

work page 1999
[66]

Lo, Albert Y. , doi =. Z. Wahrsch. Verw. Gebiete , mrclass =. 1982 , bdsk-url-1 =

work page 1982
[67]

M. Log. Scand. J. Statist. , mrclass =

work page
[68]

Statistical inference and simulation for spatial point processes , volume =

M. Statistical inference and simulation for spatial point processes , volume =

work page
[69]

Efficient nonparametric

Monard, Fran. Efficient nonparametric. 2019 , publisher=

work page 2019
[70]

and Murphy, Thomas B

Ng, Tin Lok J. and Murphy, Thomas B. , TITLE =. Canad. J. Statist. , FJOURNAL =. 2019 , NUMBER =

work page 2019
[71]

Nickl, R. , isbn=. 2023 , publisher=

work page 2023
[72]

, title =

Nyssbaum, M. , title =. Theory of Probability & Its Applications , volume =. 1987 , doi =

work page 1987
[73]

Rasmussen, Carl Edward and Williams, Christopher K. I. , title =

work page
[74]

Adaptive

Ray, Kolyan , journal=. Adaptive

work page
[75]

Rosenblatt, Murray , year=

work page
[76]

Approximate

Rue, H. Approximate. J. Roy. Statist. Soc. Ser. B , mrclass =. 2009 , bdsk-url-1 =. doi:10.1111/j.1467-9868.2008.00700.x , fjournal =

work page doi:10.1111/j.1467-9868.2008.00700.x 2009
[77]

Silverman, B. W. , TITLE =. 1986 , PAGES =

work page 1986
[78]

and van der Vaart, A

Szabo, B. and van der Vaart, A. and van Zanten, J. H. , journal=. Frequentist coverage of adaptive nonparametric

work page
[79]

2002 , publisher=

Random heterogeneous materials: microstructure and macroscopic properties , author=. 2002 , publisher=

work page 2002
[80]

van der Vaart, A. W. and van Zanten, J. H. , date-added =. Rates of contraction of posterior distributions based on. Ann. Statist. , number =

work page

Showing first 80 references.