betaFeedback

WorksMathematics

Asymptotics of Gibbs Posterior Measures on Infinite-Dimensional Hypothesis Spaces

Abstract

We prove almost sure contraction rates and consistency for Gibbs posterior measures on arbitrary hypothesis spaces, in general infinite dimensional, under dependent data, nonadditive and time-dependent losses and learning rates, misspecified models, and unattainable minimisers of the limiting objective. In particular, we develop a novel proof approach in contraction rate theory by studying the asymptotics of Gibbs measures through Lorentz spaces, real interpolation theory, and Laplace transforms dispensing with testing or variational formulations which are unavailable in the considered generality. As a consequence we obtain almost sure contraction rates for dominated Bayes and fractional Bayes posterior measures in the same generality, extending existing theory on misspecified Bayesian inference. Furthermore, we demonstrate our theoretical results on novel applications to Gibbs posteriors for Dirichlet process priors with a nonadditive Wasserstein metric loss function under ergodic, non-Markovian observables.

Introduction

The Gibbs posterior measure over any measurable space (Θ,T)(\Theta, \mathcal{T}) is a probability measures formed by combining a prior measure Π0\Pi_0 with a non-negative function RtR_t, connecting observables X1tX_1^t and hypotheses PP, typically through an exponentiated negative loss function and learning rate as Rt(X1t,P)=exp⁡(−tβtLt(X1t,P))R_t(X_1^t,P)=\exp(-t\beta_tL_t(X_1^t,P)), replacing the likelihood of Bayes posteriors,

Πt(A∣X1t)=∫ARt(X1t,P) Π0(dP)/Zt(Θ),(1)\Pi_t(A\mid X_1^t)=\int_A R_t(X_1^t,P)\,\Pi_0(\mathrm{d}P)/Z_t(\Theta), \tag*{(1)}

where Zt(Θ)Z_t(\Theta) is the normalising constant. These posterior measures, also referred to as generalised Bayes posteriors, have seen explosive interest in the last decade following the seminal work of Bissiri, Holmes, and Walker [13], who provide these as a principled posterior inference framework, mitigating misspecified models and hypothesis spaces even when Bayes fails (cf. [33]).

Precisely because the choices of loss functions and learning rates allow for more control of the objective are we required to establish that these posteriors behave desirably. However, [13] formulate their approach in the parametric setting based on a coherence axiom, whereas the literature has moved beyond this classical and parametric approach to consider non-additive loss functions, dependent observables, and infinite dimensional hypothesis spaces [4, 6, 19, 20, 42, 43, 47, 48, 52, 70]. Their theoretical study has, however, largely been restricted to rather limited settings with additive loss functions and require the existence of minimisers of the loss, or been confined to particular instantiations of the Gibbs posterior [31, 47, 50, 53, 63].

Asymptotic consistency of Gibbs posteriors is precisely what licenses that these concentrate on hypotheses that are of interest and close to optimal so that they may be used to mitigate model misspecification, and hence we cannot perform Gibbs posterior inference without understanding their asymptotics. Moreover, establishing contraction rates provides that consistency does not merely take place arbitrarily slowly but rather at a useful speed.

In this paper we will rigorously extend the applicability of Gibbs posteriors by studying the asymptotics of Gibbs posterior measures when the observables are dependent, the hypothesis space may be arbitrary and infinite-dimensional, the model misspecified, the loss function nonadditive, and no minimiser of the limiting objective may exist. To achieve this, we establish novel proof techniques in posterior asymptotics through Lorentz function spaces and real interpolation and rearrangement theory, as existing approaches to posterior asymptotics, including testing [28] and variational formulations [3, 36, 51], are generally unavailable.

The Gibbs posterior measure

Let X1,X2,…X_1,X_2,\ldots be a sequence of observables taking values in the measurable space (Ξ,X)(\Xi,\mathcal{X}) generated from the probability space (Ω,F,P0)(\Omega,\mathcal{F},P_0) where P0P_0 is the infinite process measure of the sequence (Xt)t∈N(X_t)_{t\in\mathbb{N}}. Denote a finite subsequence X1,…,Xt≡X1tX_1,\ldots,X_t \equiv X_1^t and let σ(X1t)\sigma(X_1^t) be its natural filtration.

Suppose (Θ,T)(\Theta,\mathcal{T}) is the hypothesis space, in general misspecified i.e. P0∉ΘP_0\notin\Theta, with Π0\Pi_0 a prior probability measure on this space. Let Lt:Ξt×Θ→[−∞,+∞]L_t:\Xi^t\times\Theta\to[-\infty,+\infty] be a sequence of extended real valued functions connecting observables and hypotheses. We call these loss functions although these need not be bounded below, and we naturally assume this definition extends to include P0∉ΘP_0\notin\Theta. Further, let βt>0\beta_t>0 be a deterministic sequence, and denote st:=tβts_t:=t\beta_t. We assume throughout that for all δ>0,∑te−δst<∞\delta>0, \sum_t e^{-\delta s_t}<\infty, a sufficient condition is st/log⁡t→+∞s_t/\log t\to+\infty.

The Gibbs posterior measure is thus defined as

Πt(A∣X1t)=∫Aexp⁡(−tβtLt(X1t;P))Π0(dP)Zt(Θ)(2)\Pi_t(A\mid X_1^t)=\frac{\int_A \exp(-t\beta_t L_t(X_1^t;P))\Pi_0(dP)}{Z_t(\Theta)} \tag*{(2)}

for all A∈TA\in\mathcal{T}, where

Zt(A):=∫Ae−stLt(X1t;P)Π0(dP),st:=tβt.(3)Z_t(A):=\int_A e^{-s_tL_t(X_1^t;P)}\Pi_0(dP),\qquad s_t:=t\beta_t. \tag*{(3)}

We prefer the term Gibbs posterior [17, 34] to “generalised Bayes” to illustrate that we do not necessitate the coherence of Bissiri, Holmes, and Walker [13] but rather extend their applicability and may nevertheless perform principled posterior inference. Moreover, this further highlights the intricate relationship to the use of “Gibbs measures” in PDE theory and mathematical physics (see for instance Barashkov and Gubinelli [5]).

Equivalently, (2) may be written using Δt(X1t;P)=Lt(X1t;P)−Lt(X1t;P0)\Delta_t(X_1^t;P)=L_t(X_1^t;P)-L_t(X_1^t;P_0), where we assume that Lt(X1t;P0)∈RL_t(X_1^t;P_0)\in\mathbb{R} a.s., and Rt(P):=exp⁡{−stΔt(X1t;P)}R_t(P):=\exp\{-s_t\Delta_t(X_1^t;P)\} as

Πt(A∣X1t):=∫ARt(P)Π0(dP)∫ΘRt(P)Π0(dP).(4)\Pi_t(A\mid X_1^t):=\frac{\int_A R_t(P)\Pi_0(dP)}{\int_\Theta R_t(P)\Pi_0(dP)}. \tag*{(4)}

We assume throughout for Rt:Ξ(t)×Θ→[0,∞]R_t:\Xi^{(t)}\times\Theta\to[0,\infty] to be an extended real valued function that is σ(X1t)⊗T\sigma(X_1^t)\otimes\mathcal{T} measurable for all tt and for which the denominator is a.s. positive so that the object of study is well defined.

While we work with Δt(X1t;P)\Delta_t(X_1^t;P) for familiarity of the objective in Gibbs posterior inference, we stress that this is not necessary for our results. We may likewise work with RtR_t directly as a function mapping to [0,+∞][0,+\infty], and further have this be a random sequence of functions without observables present, Rt(ω,P)R_t(\omega,P), or fully deterministic Rt(P)R_t(P).

Remark 1.1. We have not assumed any topology on any of the measurable spaces, and these are in general assumed to be infinite dimensional. The observables can be arbitrarily dependent, and we do not in general suppose that the model is well specified in that P0∈ΘP_0\in\Theta, that is we are in the MM–open case of [10]. The misspecified setting itself motivated the development of the Gibbs posteriors, and hence any such assumption would be counterproductive. Neither do we assume that there exists some common dominating measure, nor that the hypotheses are probability measures, they likewise could be (generalised) functions in some function space. The notation P∈ΘP\in\Theta has been adapted to relate to the canonical setting of the Generalised Bayes literature of utilising probability distributions in the loss, albeit in the parametric setting, and to highlight the nonparametric generality.

Remark 1.2. Our definition of Δt\Delta_t utilising P0P_0 is motivated by not requiring the existence of a minimiser of the limiting objective, which we make precise below, and hence we may be unable to work with a minimiser as in the existing literature on asymptotics of Gibbs measures [53, 63] and misspecified Bayes [39, 67].

Contraction rates for Gibbs posterior measures

Let ϵ(t)\epsilon(t) be a positive rate function converging to zero, and P↦ζ(P)P\mapsto\zeta(P) a function that quantifies the optimality of the hypotheses P∈ΘP\in\Theta in some sense which we make precise later. Consider the decreasing sequence of sets

Nt:={P∈Θ:ζ(P)≲ϵ(t)}.(5)N_t:=\{P\in\Theta:\zeta(P)\lesssim\epsilon(t)\}. \tag*{(5)}

Contraction rates ask for ϵ(t)\epsilon(t) decaying sufficiently slowly for the corresponding NtN_t to retain sufficient prior mass, while decaying sufficiently quickly for the rate to remain meaningful. In this paper we will primarily investigate sufficient, and in some cases necessary, conditions under which the following conjecture holds true.

Conjecture 1.1. For ϵ(t)→0\epsilon(t) \to0 and tϵ(t)→+∞t\epsilon(t) \to+\infty, it holds, under certain conditions on the prior hypothesis space, data generating process, and functions LtL_t, that

Πt({P∈Θ:ζ(P)≲ϵ(t)}|X1t)→1P0-a.s.(6)\Pi_t\left(\left\{P \in\Theta: \zeta(P) \lesssim\epsilon(t)\right\}\middle|X_1^t\right) \to1 \qquad P_0\text{-a.s.} \tag*{(6)}

We answer the above in the affirmative in Theorems 3.1 and 3.3 where the conditions are made precise. For now, we leave these conditions abstract.

Much of the difficulty in establishing the above conjecture lies in the considered generality of arbitrary hypothesis spaces, dependent observables, non-additive and time-dependent loss functions, and unattained infima of P↦ζ(P)P \mapsto\zeta(P).

Whereas the Bayesian setting provides natural choices for P↦ζ(P)P \mapsto\zeta(P) even under misspecification (see [39, 40]), the Gibbs posterior requires more care in identifying a sequence of functions LtL_t with desirable asymptotic properties. It turns out that the contraction rate ϵ(t)\epsilon(t) is intricately related to the rate of convergence of Δt\Delta_t to some limiting function in L1(P)L^1(P). However, it is well known that the speed of such convergence may indeed be arbitrarily slow [41]. Moreover, under non-additive and time-varying functions LtL_t, accounting also for prior misspecification where optimal parts of the hypothesis space may be outside the support of the prior, it is non-obvious what ζ(P)\zeta(P) should be.

In this paper, we utilise, adapt, and extend ideas of [58] to consider an analogue of the Kullback–Leibler neighbourhoods for Gibbs posteriors where these instead are informed by the limiting function of EP0(t)Δt\mathbb{E}_{P_0}^{(t)}\Delta_t instead. To account for model and prior misspecification, we subtract the essential infimum of this limit under the prior, even when this may not be attained.

One may further ask, in relation to the results of Miller [53], whether we require some pointwise almost sure convergence of Δt(P)\Delta_t(P) to some limiting function for asymptotic consistency or contraction rates to hold. We answer this in the negative.

Contributions

The main contributions of this work are the establishment of the first contraction rates for Gibbs posteriors for the nonparametric and arbitrary hypothesis spaces setting, nonadditive loss functions, unattained loss minimisers, misspecified models, and the almost sure case. Our results provide a novel proof technique in posterior asymptotics, building on ideas from real interpolation spaces, the Marcinkiewicz interpolation theorem, Calderón–Zygmund theory, and the Laplace transform. We demonstrate our results in novel settings that were previously unavailable to existing theory, including Dirichlet process priors, as well as observables that are not Markov at any order.

As dominated Bayes posteriors are Gibbs posteriors, our results extend to consistency and contraction rates for (fractional) Bayes under the same desired generality—dependent observables, misspecified models, arbitrary hypothesis spaces, unattained minimisers, and almost sure results.

Moreover, in the course of the development of this paper, we have explored the work of Shalizi [59] as one possible approach, but have found the theory to be incorrect. We expand on this in Appendix § 19 where we provide an explicit counterexample to the contraction rate theorem of [59].

Outline of the paper

In this part, we formally introduce the Gibbs posterior and present our main results on almost sure contraction rates of nonparametric Gibbs posteriors. § 2 develops the necessary notation, and standing assumptions, required when stating the main results on contraction rates and posterior consistency in § 3. This includes their relation to the Bayes posterior, and demonstrating applicability of our results to tempered Bayes posterior inference in the same generality. We further present an example of our theory on Dirichlet process priors under dependent observables. To contextualise the need for novel proof approaches, we discuss the existing theoretical work on posterior asymptotics in § 5, including why previous proof approaches in Bayesian nonparametric contraction rates are unavailable in the considered setting. In § 6, we outline the rather lengthy proof arguments and explain the main ideas behind these. The full proofs can be found in Parts I to III. A discussion on the main results and open problems follows in § 7. Moreover, during the completion of this paper, we have found the results of Shalizi [59] to be incorrect, and provide an explicit counterexample and a discussion on this in Appendix § 19.

Acknowledgements

The author is indebted to Geoff Nicholls for his guidance and support, and for providing invaluable suggestions and corrections to an earlier draft of the paper. Further, the author acknowledges support from an EPSRC Industrial Doctoral Landscape Award [UKRI2002] and GSK.

Notation, and standing assumptions

In the following arguments, for positive integers l,r∈N:={1,2,…}l,r \in\mathbb{N} := \{1,2,\ldots\}, l≤rl \le r we denote [l;r]:={l,l+1,…,r}[l;r] := \{l,l+1,\ldots,r\}. Further, we say at≳bta_t \gtrsim b_t eventually if there exists some unspecified generic constant c>0c > 0 such that at≥c⋅bta_t \ge c \cdot b_t for all tt sufficiently large, and denote at∼bta_t \sim b_t iff at≳bta_t \gtrsim b_t and bt≳atb_t \gtrsim a_t. The maximum is denoted a∨b=max⁡(a,b)a \vee b = \max(a,b) and the minimum a∧b=min⁡(a,b)a \wedge b = \min(a,b). The complement of a set A∈TA \in\mathcal{T} is denoted Ac:=Θ∖AA^{c} := \Theta\setminus A.

To avoid confusion, we say a function converges P0P_0–almost surely when referring to (Ω,P0)(\Omega,P_0) and Π0\Pi_0–almost everywhere when meaning (Θ,Π0)(\Theta,\Pi_0).

General definitions

It will be convenient to consider, instead of the classical Lebesgue function spaces, a generalisation of these which will allow us to study the types of functions for which the Gibbs posterior is consistent and the corresponding contraction rates more accurately.

Definition 2.1. We denote Lorentz spaces on a σ\sigma–finite measure space (X,μ)(X,\mu) by Lp,∞(X,μ)L^{p,\infty}(X,\mu), also known as weak LpL^p spaces, which, for p∈[1,∞)p \in[1,\infty) are equipped with the Lorentz quasi–norm

∥u∥Lp,∞(μ)=sup⁡λ>0λ⋅[μ{x∈X:∣u(x)∣>λ}]1/p.(7)\lVert u \rVert_{L^{p,\infty}(\mu)} = \sup_{\lambda>0} \lambda\cdot\left[\mu\{x \in X : |u(x)| > \lambda\}\right]^{1/p}. \tag*{(7)}

For p=∞p = \infty the Lorentz space is naturally identified with the L∞L^\infty space.1

References

  1. [1]Paul H. Algoet and Thomas M. Cover, A sandwich proof of the Shannon-McMillan-Breiman theorem, Ann. Probab. 16 (1988), no. 2, 899–909. MR929085DOI
  2. [2]Pierre Alquier and James Ridgway, Concentration of tempered posteriors and of their variational approximations, Ann. Statist. 48 (2020), no. 3, 1475–1497. MR4124331arxiv.org/abs/1706.09293
  3. [3]Pierre Alquier, James Ridgway, and Nicolas Chopin, On the properties of variational approximations of Gibbs posteriors, J. Mach. Learn. Res. 17 (2016), Paper No. 239, 41. MR3595173arxiv.org/abs/1506.04091
  4. [4]Matias Altamirano, Francois-Xavier Briol, and Jeremias Knoblauch, Robust and conjugate Gaussian process regression, Proc. 41st Int. Conf. Mach. Learn. (ICML), 2024, pp. 1155–1185. arXiv:2311.00463.arxiv.org/abs/2311.00463
  5. [5]N. Barashkov and M. Gubinelli, A variational method for Φ⁴₃, Duke Math. J. 169 (2020), no. 17, 3339–3415. MR4173157
  6. [6]Yannick Baraud, From robust tests to Bayes-like posterior distributions, Probab. Theory Related Fields 188 (2024), no. 1-2, 159–234. MR4685697arxiv.org/abs/2107.12011
  7. [7]Andrew Barron, Mark J. Schervish, and Larry Wasserman, The consistency of posterior distributions in nonparametric problems, Ann. Statist. 27 (1999), no. 2, 536–561. MR1714718DOI
  8. [8]Colin Bennett and Robert Sharpley, Interpolation of operators, Pure and Applied Mathematics, vol. 129, Academic Press, Inc., Boston, MA, 1988. MR0928802
  9. [9]Robert H. Berk, Limiting behavior of posterior distributions when the model is incorrect, Ann. Math. Statist. 37 (1966), 51–58; correction, ibid. 745–746. MR189176DOI
  10. [10]Jose-M. Bernardo and Adrian F. M. Smith, Bayesian theory, Wiley Series in Probability and Mathematical Statistics: Probability and Mathematical Statistics, John Wiley & Sons, Ltd., Chichester, 1994. MR1274699
  11. [11]Anirban Bhattacharya, Debdeep Pati, and Yun Yang, Bayesian fractional posteriors, Ann. Statist. 47 (2019), no. 1, 39–66. MR3909926arxiv.org/abs/1611.01125
  12. [12]Patrick Billingsley, Convergence of probability measures, 2nd ed., Wiley, New York, 1999 (English). MR1700749
  13. [13]P. G. Bissiri, C. C. Holmes, and S. G. Walker, A general framework for updating belief distributions, J. R. Statist. Soc. Ser. B Statist. Methodol. 78 (2016), no. 5, 1103–1130. MR3557191DOI
  14. [14]A. P. Calderón, Spaces between L¹ and L∞ and the theorem of Marcinkiewicz, Studia Math. 26 (1966), 273–299. MR203444DOI
  15. [15]A. P. Calderón and A. Zygmund, On the existence of certain singular integrals, Acta Math. 88 (1952), 85–139. MR52553
  16. [16]F. Camerlenghi, E. Dolera, S. Favaro, and E. Mainini, Wasserstein posterior contraction rates in non-dominated Bayesian nonparametric models, Ann. Inst. Henri Poincaré Probab. Stat. 62 (2026), no. 1, 582–606. MR5036538arxiv.org/abs/2201.12225
  17. [17]Olivier Catoni, Pac-Bayesian supervised classification: the thermodynamics of statistical learning, Institute of Mathematical Statistics Lecture Notes—Monograph Series, vol. 56, Institute of Mathematical Statistics, Beachwood, OH, 2007. MR2483528DOI
  18. [18]Douglas Coates, Ian Melbourne, and Amin Talebi, Natural measures and statistical properties of non-statistical maps with multiple neutral fixed points, J. Lond. Math. Soc. (2) 113 (2026), no. 5, Paper No. e70569, 32. MR5077628arxiv.org/abs/2407.07286
  19. [19]Charita Dellaporta and Theodoros Damoulas, Robust Bayesian inference for measurement error misspecification: the Berkson and classical cases, Electron. J. Statist. 20 (2026), no. 1, 445–502. MR5027587arxiv.org/abs/2306.01468
  20. [20]Charita Dellaporta, Jeremias Knoblauch, Theodoros Damoulas, and Francois-Xavier Briol, Robust bayesian inference for simulator-based models via the mmd posterior bootstrap, Proc. 25th Int. Conf. Artif. Intell. Statist., 2022, pp. 943–970. arXiv:2202.04744.arxiv.org/abs/2202.04744
  21. [21]P. Diaconis and D. Freedman, On inconsistent Bayes estimates of location, Ann. Statist. 14 (1986), no. 1, 68–87. MR829556DOI
  22. [22]Emanuele Dolera, Stefano Favaro, and Edoardo Mainini, Strong posterior contraction rates via Wasserstein dynamics, Probab. Theory Related Fields 189 (2024), no. 1-2, 659–720. MR4750566arxiv.org/abs/2203.10754
  23. [23]Thomas S. Ferguson, Prior distributions on spaces of probability measures, Ann. Statist. 2 (1974), 615–629. MR438568DOI
  24. [24]Jonathan M. Fraser, Assouad dimension and fractal geometry, Cambridge Tracts in Mathematics, vol. 222, Cambridge University Press, Cambridge, 2021. MR4411274arxiv.org/abs/2005.03763
  25. [25]David Freedman, On the Bernstein-von Mises theorem with infinite-dimensional parameters, Ann. Statist. 27 (1999), no. 4, 1119–1140. MR1740119
  26. [26]Subhashis Ghosal, Jayanta K. Ghosh, and Aad W. van der Vaart, Convergence rates of posterior distributions, Ann. Statist. 28 (2000), no. 2, 500–531. MR1790007
  27. [27]Subhashis Ghosal and Aad W. van der Vaart, Convergence rates of posterior distributions for non-i.i.d. observations, Ann. Statist. 35 (2007), no. 1, 192–223. MR2332274
  28. [28]———, Fundamentals of nonparametric Bayesian inference, Cambridge Series in Statistical and Probabilistic Mathematics, vol. 44, Cambridge University Press, Cambridge, 2017. MR3587782DOI
  29. [29]Matteo Giordano and Hanne Kekkonen, Bernstein–von Mises theorems and uncertainty quantification for linear inverse problems, SIAM/ASA J. Uncertain. Quantif. 8 (2020), no. 1, 342–373. MR4069334arxiv.org/abs/1811.04058
  30. [30]Loukas Grafakos, Classical Fourier analysis, Third, Graduate Texts in Mathematics, vol. 249, Springer, New York, 2014. MR3243734
  31. [31]Peter D. Grünwald and Nishant A. Mehta, Fast rates for general unbounded loss functions: from ERM to generalized Bayes, J. Mach. Learn. Res. 21 (2020), Paper No. 56, 80. MR4095335
  32. [32]Peter D. Grünwald and Thijs van Ommen, Inconsistency of Bayesian inference for misspecified linear models, and a proposal for repairing it, Bayesian Anal. 12 (2017), no. 4, 1069–1103. MR3724979arxiv.org/abs/1412.3730
  33. [33]Jack Jewson, Jim Q. Smith, and Chris Holmes, On the stability of general Bayesian inference, Bayesian Anal. 21 (2026), no. 2, 725–755. MR5086491
  34. [34]Wenxin Jiang and Martin A. Tanner, Gibbs posterior for variable selection in high-dimensional classification and data mining, Ann. Statist. 36 (2008), no. 5, 2207–2231. MR2458185DOI
  35. [35]Yuji Kasahara and Nobuko Kosugi, Remarks on Tauberian theorem of exponential type and Fenchel-Legendre transform, Osaka J. Math. 39 (2002), no. 3, 613–619. MR1932284DOI
  36. [36]Anya Katsevich and Philippe Rigollet, On the approximation accuracy of Gaussian variational inference, Ann. Statist. 52 (2024), no. 4, 1384–1409. MR4804813arxiv.org/abs/2301.02168
  37. [37]David Kerr and Hanfeng Li, Ergodic theory, Springer Monographs in Mathematics, Springer, Cham, 2016. Independence and dichotomies. MR3616077
  38. [38]B. J. K. Kleijn, Frequentist validity of Bayesian limits, Ann. Statist. 49 (2021), no. 1, 182–202. MR4206674arxiv.org/abs/1611.08444
  39. [39]B. J. K. Kleijn and A. W. van der Vaart, Misspecification in infinite-dimensional Bayesian statistics, Ann. Statist. 34 (2006), no. 2, 837–877. MR2283395DOI
  40. [40]———, The Bernstein-Von-Mises theorem under misspecification, Electron. J. Statist. 6 (2012), 354–381. MR2988412DOI
  41. [41]Ulrich Krengel, On the speed of convergence in the ergodic theorem, Monatsh. Math. 86 (1978), no. 1, 3–6. MR510630DOI
  42. [42]William Laplante, Matias Altamirano, Andrew B. Duncan, Jeremias Knoblauch, and François-Xavier Briol, Robust and conjugate spatio-temporal Gaussian processes, Proc. 42nd Int. Conf. Mach. Learn. (ICML), 2025, pp. 32562–32592. arXiv:2502.02450.arxiv.org/abs/2502.02450
  43. [43]Jeong Eun Lee, Sitong Liu, and Geoff K. Nicholls, Bayesian inference for the learning rate in generalised Bayesian inference, 2026. preprint arXiv:2506.12532.arxiv.org/abs/2506.12532
  44. [44]Alice L’Huillier, Luke Travis, Ismaël Castillo, and Kolyan Ray, Semiparametric inference using fractional posteriors, J. Mach. Learn. Res. 24 (2023), Paper No. [389], 61. MR4720845arxiv.org/abs/2301.08158
  45. [45]Elliott H. Lieb and Michael Loss, Analysis, Graduate Studies in Mathematics, vol. 14, American Mathematical Society, Providence, RI, 1997. MR1415616
  46. [46]Artur O. Lopes, Silvia R. C. Lopes, and Paulo Varandas, Bayes posterior convergence for loss functions via almost additive thermodynamic formalism, J. Stat. Phys. 186 (2022), no. 3, Paper No. 35, 34. MR4375840arxiv.org/abs/2012.05601
  47. [47]Takuo Matsubara, Jeremias Knoblauch, François-Xavier Briol, and Chris J. Oates, Robust generalised Bayesian inference for intractable likelihoods, J. R. Statist. Soc. Ser. B. Statist. Methodol. 84 (2022), no. 3, 997–1022. MR4460583arxiv.org/abs/2104.07359
  48. [48]———, Generalized Bayesian inference for discrete intractable likelihood, J. Amer. Statist. Assoc. 119 (2024), no. 547, 2345–2355. MR4797945
  49. [49]Pertti Mattila, Rectifiability—a survey, London Mathematical Society Lecture Note Series, vol. 483, Cambridge University Press, Cambridge, 2023. MR4520153
  50. [50]Kevin McGoff, Sayan Mukherjee, and Andrew B. Nobel, Gibbs posterior convergence and the thermodynamic formalism, Ann. Appl. Probab. 32 (2022), no. 1, 461–496. MR4386533arxiv.org/abs/1901.08641
  51. [51]Terje Mildner, Paris Giampouras, and Theodoros Damoulas, Rates of convergence of generalised variational inference posteriors under prior misspecification, 2025. preprint arXiv:2510.03109.arxiv.org/abs/2510.03109
  52. [52]Terje Mildner, Oliver Hamelijnck, Paris Giampouras, and Theodoros Damoulas, Federated generalised variational inference: A robust probabilistic federated learning framework, Proc. 42nd Int. Conf. Mach. Learn. (ICML), 2025, pp. 44134–44174. arXiv:2502.00846.arxiv.org/abs/2502.00846
  53. [53]Jeffrey W. Miller, Asymptotic normality, concentration, and coverage of generalized posteriors, J. Mach. Learn. Res. 22 (2021), Paper No. 168, 53. MR4318524
  54. [54]XuanLong Nguyen, Convergence of latent mixing measures in finite and infinite mixture models, Ann. Statist. 41 (2013), no. 1, 370–400. MR3059422DOI
  55. [55]Richard Nickl, Bernstein–von Mises theorems for statistical inverse problems I: Schrödinger equation, J. Eur. Math. Soc. (JEMS) 22 (2020), no. 8, 2697–2750. MR4118619arxiv.org/abs/1707.01764
  56. [56]Judith Rousseau and Catia Scricciolo, Wasserstein convergence in Bayesian and frequentist deconvolution models, Ann. Statist. 52 (2024), no. 4, 1691–1715. MR4804824DOI
  57. [57]Walter Rudin, Real and complex analysis, 3rd ed., McGraw-Hill, Inc., USA, 1987. MR924157
  58. [58]Lorraine Schwartz, On Bayes procedures, Z. Wahrscheinlichkeitstheorie und Verw. Gebiete 4 (1965), 10–26. MR184378pubmed.ncbi.nlm.nih.gov/16591192
  59. [59]Cosma Rohilla Shalizi, Dynamics of Bayesian updating with dependent data and misspecified models, Electron. J. Statist. 3 (2009), 1039–1074. MR2557128DOI
  60. [60]Xiaotong Shen and Larry Wasserman, Rates of convergence of posterior distributions, Ann. Statist. 29 (2001), no. 3, 687–714. MR1865337DOI
  61. [61]A. M. Stuart, Inverse problems: a Bayesian perspective, Acta Numer. 19 (2010), 451–559. MR2652785DOI
  62. [62]Nicholas Syring and Ryan Martin, Robust and rate-optimal Gibbs posterior inference on the boundary of a noisy image, Ann. Statist. 48 (2020), no. 3, 1498–1513. MR4124332DOI
  63. [63]———, Gibbs posterior concentration rates under sub-exponential type losses, Bernoulli 29 (2023), no. 2, 1080–1108. MR4550216DOI
  64. [64]A. W. van der Vaart and J. H. van Zanten, Rates of contraction of posterior distributions based on Gaussian process priors, Ann. Statist. 36 (2008), no. 3, 1435–1463. MR2418663DOI
  65. [65]Cédric Villani, Optimal transport: old and new, Grundlehren der mathematischen Wissenschaften [Fundamental Principles of Mathematical Sciences], vol. 338, Springer-Verlag, Berlin, 2009. MR2459454
  66. [66]Stephen G. Walker, New approaches to Bayesian consistency, Ann. Statist. 32 (2004), no. 5, 2028–2043. MR2102501DOI
  67. [67]———, Bayesian inference with misspecified models, J. Statist. Plann. Inference 143 (2013), no. 10, 1621–1633. MR3088220DOI
  68. [68]Stephen G. Walker, Antonio Lijoi, and Igor Prünster, On rates of convergence for posterior distributions in infinite-dimensional models, Ann. Statist. 35 (2007), no. 2, 738–746. MR2336866DOI
  69. [69]Benjamin Weiss, Subshifts of finite type and sofic systems, Monatsh. Math. 77 (1973), 462–474. MR340556DOI
  70. [70]Steven Winter, Omar Meliketchi, and David B. Dunson, Sequential Gibbs posteriors with applications to principal component analysis, Biometrika 113 (2026), no. 2, Paper No. asag020, 20. MR5081051arxiv.org/abs/2310.12882
  71. [71]Arnold Zellner, Optimal information processing and Bayes’s theorem, Amer. Statist. 42 (1988), no. 4, 278–284. With comments and a reply by the author. MR971095
  72. [72]Tong Zhang, From ε-entropy to KL-entropy: analysis of minimum information complexity density estimation, Ann. Statist. 34 (2006), no. 5, 2180–2210. MR2291497DOI

Paper details

vlnc:W_hg7xnt9w
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents