arXiv subjects

Christopher Drovandi

Publications and source records attributed to Christopher Drovandi.

At least 37 records · Page 2Linked to original sources

arXiv2025

Model Selection for Gaussian-gated Gaussian Mixture of Experts Using Dendrograms of Mixing Measures

Mixture of Experts (MoE) models constitute a widely utilized class of ensemble learning approaches in statistics and machine learning, known for their flexibility and computational efficiency. They have become integral components in numerous state-of-the-art deep neural network architectures, particularly for analyzing heterogeneous data across diverse domains. Despite their practical success, the theoretical understanding of model selection, especially concerning the optimal number of mixture components or experts, remains limited and poses significant challenges. These challenges primarily stem from the inclusion of covariates in both the Gaussian gating functions and expert networks, which introduces intrinsic interactions governed by partial differential equations with respect to their parameters. In this paper, we revisit the concept of dendrograms of mixing measures and introduce a novel extension to Gaussian-gated Gaussian MoE models that enables consistent estimation of the true number of mixture components and achieves the pointwise optimal convergence rate for parameter estimation in overfitted scenarios. Notably, this approach circumvents the need to train and compare a range of models with varying numbers of components, thereby alleviating the computational burden, particularly in high-dimensional or deep neural network settings. Experimental results on synthetic data demonstrate the effectiveness of the proposed method in accurately recovering the number of experts. It outperforms common criteria such as the Akaike information criterion, the Bayesian information criterion, and the integrated completed likelihood, while achieving optimal convergence rates for parameter estimation and accurately approximating the regression function.

stat.ML↗

arXiv2025

Robustifying Approximate Bayesian Computation

Approximate Bayesian computation (ABC) is one of the most popular "likelihood-free" methods. These methods have been applied in a wide range of fields by providing solutions to intractable likelihood problems in which exact Bayesian approaches are either infeasible or computationally costly. However, the performance of ABC can be unreliable when dealing with model misspecification. To circumvent the poor behavior of ABC in these settings, we propose a novel ABC approach that is robust to model misspecification. This new method can deliver more accurate statistical inference under model misspecification than alternatives and also enables the detection of summary statistics that are incompatible with the assumed data-generating process. We demonstrate the effectiveness of our approach through several simulated examples, where it delivers more accurate point estimates and uncertainty quantification over standard ABC approaches when the model is misspecified. Additionally, we apply our approach to an empirical example, further showcasing its advantages over alternative methods.

stat.ME↗

arXiv2025

Unbiased and Consistent Nested Sampling via Sequential Monte Carlo

We introduce a new class of sequential Monte Carlo methods which reformulates the essence of the nested sampling method of Skilling (2006) in terms of sequential Monte Carlo techniques. Two new algorithms are proposed, nested sampling via sequential Monte Carlo (NS-SMC) and adaptive nested sampling via sequential Monte Carlo (ANS-SMC). The new framework allows convergence results to be obtained in the setting when Markov chain Monte Carlo (MCMC) is used to produce new samples. An additional benefit is that marginal likelihood (normalising constant) estimates given by NS-SMC are unbiased. In contrast to NS, the analysis of our proposed algorithms does not require the (unrealistic) assumption that the simulated samples be independent. We show that a minor adjustment to our ANS-SMC algorithm recovers the original NS algorithm, which provides insights as to why NS seems to produce accurate estimates despite a typical violation of its assumptions. A numerical study is conducted where the performance of the proposed algorithms and temperature-annealed SMC is compared on challenging problems. Code for the experiments is made available online at https://github.com/LeahPrice/SMC-NS .

stat.CO↗

arXiv2025

Simulation-based Bayesian inference under model misspecification

Simulation-based Bayesian inference (SBI) methods are widely used for parameter estimation in complex models where evaluating the likelihood is challenging but generating simulations is relatively straightforward. However, these methods commonly assume that the simulation model accurately reflects the true data-generating process, an assumption that is frequently violated in realistic scenarios. In this paper, we focus on the challenges faced by SBI methods under model misspecification. We consolidate recent research aimed at mitigating the effects of misspecification, highlighting three key strategies: i) robust summary statistics, ii) generalised Bayesian inference, and iii) error modelling and adjustment parameters. To illustrate both the vulnerabilities of popular SBI methods and the effectiveness of misspecification-robust alternatives, we present empirical results on an illustrative example.

stat.ME↗

arXiv2024

The Statistical Accuracy of Neural Posterior and Likelihood Estimation

Neural posterior estimation (NPE) and neural likelihood estimation (NLE) are machine learning approaches that provide accurate posterior, and likelihood, approximations in complex modeling scenarios, and in situations where conducting amortized inference is a necessity. While such methods have shown significant promise across a range of diverse scientific applications, the statistical accuracy of these methods is so far unexplored. In this manuscript, we give, for the first time, an in-depth exploration on the statistical behavior of NPE and NLE. We prove that these methods have similar theoretical guarantees to common statistical methods like approximate Bayesian computation (ABC) and Bayesian synthetic likelihood (BSL). While NPE and NLE methods are just as accurate as ABC and BSL, we prove that this accuracy can often be achieved at a vastly reduced computational cost, and will therefore deliver more attractive approximations than ABC and BSL in certain problems. We verify our results theoretically and in several examples from the literature.

stat.ML↗

arXiv2024

A Comprehensive Guide to Simulation-based Inference in Computational Biology

Computational models are invaluable in capturing the complexities of real-world biological processes. Yet, the selection of appropriate algorithms for inference tasks, especially when dealing with real-world observational data, remains a challenging and underexplored area. This gap has spurred the development of various parameter estimation algorithms, particularly within the realm of Simulation-Based Inference (SBI), such as neural and statistical SBI methods. Limited research exists on how to make informed choices on SBI methods when faced with real-world data, which often results in some form of model misspecification. In this paper, we provide comprehensive guidelines for deciding between SBI approaches for complex biological models. We apply the guidelines to two agent-based models that describe cellular dynamics using real-world data. Our study unveils a critical insight: while neural SBI methods demand significantly fewer simulations for inference results, they tend to yield biased estimations, a trend persistent even with robust variants of these algorithms. On the other hand, the accuracy of statistical SBI methods enhances substantially as the number of simulations increases. This finding suggests that, given a sufficient computational budget, statistical SBI can surpass neural SBI in performance. Our results not only shed light on the efficacy of different SBI methodologies in real-world scenarios but also suggest potential avenues for enhancing neural SBI approaches. This study is poised to be a useful resource for computational biologists navigating the intricate landscape of SBI in biological modeling.

stat.AP↗

arXiv2024

Ecosystem knowledge should replace coexistence and stability assumptions in ecological network modelling

Quantitative population modelling is an invaluable tool for identifying the cascading effects of ecosystem management and interventions. Ecosystem models are often constructed by assuming stability and coexistence in ecological communities as a proxy for abundance data when monitoring programs are not available. However, a growing body of literature suggests that these assumptions are inappropriate for modelling conservation outcomes. In this work, we develop an alternative for dataless population modelling that instead relies on expert-elicited knowledge of species abundances. While time series abundance data is often not available for ecosystems of interest, these systems may still be highly studied or observed in an informal capacity. In particular, limits on population sizes and their capacity to rapidly change during an observation period can be reasonably elicited for many species. We propose a robust framework for generating an ensemble of ecosystem models whose population predictions match the expected population dynamics, as defined by experts. Our new Bayesian algorithm systematically removes model parameters that lead to unreasonable population predictions without incurring excessive computational costs. Our results demonstrate that models constructed using expert-elicited information, rather than stability and coexistence assumptions, can dramatically impact population predictions, expected responses to management, conservation decision-making, and long-term ecosystem behaviour. In the absence of data, we argue that field observations and expert knowledge are preferred for representing ecosystems observed in nature instead of theoretical assumptions of coexistence and stability.

q-bio.PE↗

arXiv2024

Wasserstein Gaussianization and Efficient Variational Bayes for Robust Bayesian Synthetic Likelihood

The Bayesian Synthetic Likelihood (BSL) method is a widely-used tool for likelihood-free Bayesian inference. This method assumes that some summary statistics are normally distributed, which can be incorrect in many applications. We propose a transformation, called the Wasserstein Gaussianization transformation, that uses a Wasserstein gradient flow to approximately transform the distribution of the summary statistics into a Gaussian distribution. BSL also implicitly requires compatibility between simulated summary statistics under the working model and the observed summary statistics. A robust BSL variant which achieves this has been developed in the recent literature. We combine the Wasserstein Gaussianization transformation with robust BSL, and an efficient Variational Bayes procedure for posterior approximation, to develop a highly efficient and reliable approximate Bayesian inference method for likelihood-free problems.

stat.CO↗

arXiv2024

Preconditioned Neural Posterior Estimation for Likelihood-free Inference

Simulation based inference (SBI) methods enable the estimation of posterior distributions when the likelihood function is intractable, but where model simulation is feasible. Popular neural approaches to SBI are the neural posterior estimator (NPE) and its sequential version (SNPE). These methods can outperform statistical SBI approaches such as approximate Bayesian computation (ABC), particularly for relatively small numbers of model simulations. However, we show in this paper that the NPE methods are not guaranteed to be highly accurate, even on problems with low dimension. In such settings the posterior cannot be accurately trained over the prior predictive space, and even the sequential extension remains sub-optimal. To overcome this, we propose preconditioned NPE (PNPE) and its sequential version (PSNPE), which uses a short run of ABC to effectively eliminate regions of parameter space that produce large discrepancy between simulations and data and allow the posterior emulator to be more accurately trained. We present comprehensive empirical evidence that this melding of neural and statistical SBI methods improves performance over a range of examples, including a motivating example involving a complex agent-based model applied to real tumour growth data.

stat.ML↗

arXiv2024

Innovative Approaches in Soil Carbon Sequestration Modelling for Better Prediction with Limited Data

Soil carbon accounting and prediction play a key role in building decision support systems for land managers selling carbon credits, in the spirit of the Paris and Kyoto protocol agreements. Land managers typically rely on computationally complex models fit using sparse datasets to make these accounts and predictions. The model complexity and sparsity of the data can lead to over-fitting, leading to inaccurate results when making predictions with new data. Modellers address over-fitting by simplifying their models and reducing the number of parameters, and in the current context this could involve neglecting some soil organic carbon (SOC) components. In this study, we introduce two novel SOC models and a new RothC-like model and investigate how the SOC components and complexity of the SOC models affect the SOC prediction in the presence of small and sparse time series data. We develop model selection methods that can identify the soil carbon model with the best predictive performance, in light of the available data. Through this analysis we reveal that commonly used complex soil carbon models can over-fit in the presence of sparse time series data, and our simpler models can produce more accurate predictions. The published version of this study is available in Scientific Reports (https://www.nature.com/articles/s41598-024-53516-z/<10.1038/s41598-024-53516-z>)

stat.CO↗

arXiv2024

Unlocking ensemble ecosystem modelling for large and complex networks

The potential effects of conservation actions on threatened species can be predicted using ensemble ecosystem models by forecasting populations with and without intervention. These model ensembles commonly assume stable coexistence of species in the absence of available data. However, existing ensemble-generation methods become computationally inefficient as the size of the ecosystem network increases, preventing larger networks from being studied. We present a novel sequential Monte Carlo sampling approach for ensemble generation that is orders of magnitude faster than existing approaches. We demonstrate that the methods produce equivalent parameter inferences, model predictions, and tightly constrained parameter combinations using a novel sensitivity analysis method. For one case study, we demonstrate a speed-up from 108 days to 6 hours, while maintaining equivalent ensembles. Additionally, we demonstrate how to identify the parameter combinations that strongly drive feasibility and stability, drawing ecological insight from the ensembles. Now, for the first time, larger and more realistic networks can be practically simulated and analysed.

q-bio.PE↗

arXiv2023

Adaptively switching between a particle marginal Metropolis-Hastings and a particle Gibbs kernel in SMC$^2$

Sequential Monte Carlo squared (SMC$^2$; Chopin et al., 2012) methods can be used to sample from the exact posterior distribution of intractable likelihood state space models. These methods are the SMC analogue to particle Markov chain Monte Carlo (MCMC; Andrieu et al., 2010) and rely on particle MCMC kernels to mutate the particles at each iteration. Two options for the particle MCMC kernels are particle marginal Metropolis-Hastings (PMMH) and particle Gibbs (PG). We introduce a method to adaptively select the particle MCMC kernel at each iteration of SMC$^2$, with a particular focus on switching between a PMMH and PG kernel. The resulting method can significantly improve the efficiency of SMC$^2$ compared to using a fixed particle MCMC kernel throughout the algorithm. Code for our methods is available at https://github.com/imkebotha/kernel_switching_smc2.

stat.CO↗

arXiv2023

Many-levelled continuation ratio models for frequency of alcohol and drug use data

Studies of alcohol and drug use are often interested in the number of days that people use the substance of interest over an interval, such as 28 days before a survey date. Although count models are often used for this purpose, they are not strictly appropriate for this type of data because the response variable is bounded above. Furthermore, if some peoples' substance use behaviors are characterized by various weekly patterns of use, summaries of substance days-of-use used over longer periods can exhibit multiple modes. These characteristics of substance days-of-use data are not easily fitted with conventional parametric model families. We propose a continuation ratio ordinal model for substance days-of-use data. Instead of grouping the set of possible response values into a small set of ordinal categories, each possible value is assigned its own category. This allows the exact numeric distribution implied by the predicted ordinal response to be recovered. We demonstrate the proposed model using survey data reporting days of alcohol use over 28-day intervals. We show the continuation ratio model is better able to capture the complexity in the drinking days dataset compared to binomial, hurdle-negative binomial and beta-binomial models.

stat.AP↗

arXiv2023

Bayesian inference for misspecified generative models

Bayesian inference is a powerful tool for combining information in complex settings, a task of increasing importance in modern applications. However, Bayesian inference with a flawed model can produce unreliable conclusions. This review discusses approaches to performing Bayesian inference when the model is misspecified, where by misspecified we mean that the analyst is unwilling to act as if the model is correct. Much has been written about this topic, and in most cases we do not believe that a conventional Bayesian analysis is meaningful when there is serious model misspecification. Nevertheless, in some cases it is possible to use a well-specified model to give meaning to a Bayesian analysis of a misspecified model and we will focus on such cases. Three main classes of methods are discussed - restricted likelihood methods, which use a model based on a non-sufficient summary of the original data; modular inference methods which use a model constructed from coupled submodels and some of the submodels are correctly specified; and the use of a reference model to construct a projected posterior or predictive distribution for a simplified model considered to be useful for prediction or interpretation.

stat.ME↗

arXiv2023

Generalised likelihood profiles for models with intractable likelihoods

Likelihood profiling is an efficient and powerful frequentist approach for parameter estimation, uncertainty quantification and practical identifiablity analysis. Unfortunately, these methods cannot be easily applied for stochastic models without a tractable likelihood function. Such models are typical in many fields of science, rendering these classical approaches impractical in these settings. To address this limitation, we develop a new approach to generalising the methods of likelihood profiling for situations when the likelihood cannot be evaluated but stochastic simulations of the assumed data generating process are possible. Our approach is based upon recasting developments from generalised Bayesian inference into a frequentist setting. We derive a method for constructing generalised likelihood profiles and calibrating these profiles to achieve desired frequentist coverage for a given coverage level. We demonstrate the performance of our method on realistic examples from the literature and highlight the capability of our approach for the purpose of practical identifability analysis for models with intractable likelihoods.

stat.ME↗

arXiv2023

Bayesian Synthetic Likelihood

Bayesian statistics is concerned with conducting posterior inference for the unknown quantities in a given statistical model. Conventional Bayesian inference requires the specification of a probabilistic model for the observed data, and the construction of the resulting likelihood function. However, sometimes the model is so complicated that evaluation of the likelihood is infeasible, which renders exact Bayesian inference impossible. Bayesian synthetic likelihood (BSL) is a posterior approximation procedure that can be used to conduct inference in situations where the likelihood is intractable, but where simulation from the model is straightforward. In this entry, we give a high-level presentation of BSL, and its extensions aimed at delivering scalable and robust posterior inferences.

stat.ME↗

arXiv2023

Transport Reversible Jump Proposals

Reversible jump Markov chain Monte Carlo (RJMCMC) proposals that achieve reasonable acceptance rates and mixing are notoriously difficult to design in most applications. Inspired by recent advances in deep neural network-based normalizing flows and density estimation, we demonstrate an approach to enhance the efficiency of RJMCMC sampling by performing transdimensional jumps involving reference distributions. In contrast to other RJMCMC proposals, the proposed method is the first to apply a non-linear transport-based approach to construct efficient proposals between models with complicated dependency structures. It is shown that, in the setting where exact transports are used, our RJMCMC proposals have the desirable property that the acceptance probability depends only on the model probabilities. Numerical experiments demonstrate the efficacy of the approach.

stat.CO↗

arXiv2023

Reliable Bayesian Inference in Misspecified Models

We provide a general solution to a fundamental open problem in Bayesian inference, namely poor uncertainty quantification, from a frequency standpoint, of Bayesian methods in misspecified models. While existing solutions are based on explicit Gaussian approximations of the posterior, or computationally onerous post-processing procedures, we demonstrate that correct uncertainty quantification can be achieved by replacing the usual posterior with an intuitive approximate posterior. Critically, our solution is applicable to likelihood-based, and generalized, posteriors as well as cases where the likelihood is intractable and must be estimated. We formally demonstrate the reliable uncertainty quantification of our proposed approach, and show that valid uncertainty quantification is not an asymptotic result but occurs even in small samples. We illustrate this approach through a range of examples, including linear, and generalized, mixed effects models.

stat.ME↗