The HTI Tools Lab produces a large body of academic research as part of its work, which is published in various academic journals. See below for a full list.
THEORY AND METHODS
A Beta Cauchy-Cauchy (BECCA) shrinkage prior for Bayesian variable selection
JANUARY 2025
This paper introduces a novel Bayesian approach for variable selection in high-dimensional and potentially sparse regression settings. Our method replaces the indicator variables in the traditional spike and slab prior with continuous, Beta-distributed random variables and places half Cauchy priors over the parameters of the Beta distribution, which significantly improves the predictive and inferential performance of the technique. Similar to shrinkage methods, our continuous parameterization of the spike and slab prior enables us explore the posterior distributions of interest using fast gradient-based methods, such as Hamiltonian Monte Carlo (HMC), while at the same time explicitly allowing for variable selection in a principled framework. We study the frequentist properties of our model via simulation and show that our technique outperforms the latest Bayesian variable selection methods in both linear and logistic regression. The efficacy, applicability and performance of our approach, are further underscored through its implementation on real datasets.
arXiv: A Beta Cauchy-Cauchy (BECCA) shrinkage prior for Bayesian variable selection
Covariate Dependent Mixture of Bayesian Networks
JANUARY 2025
Learning the structure of Bayesian networks from data provides insights into underlying processes and the causal relationships that generate the data, but its usefulness depends on the homogeneity of the data population, a condition often violated in real-world applications. In such cases, using a single network structure for inference can be misleading, as it may not capture sub-population differences. To address this, we propose a novel approach of modelling a mixture of Bayesian networks where component probabilities depend on individual characteristics. Our method identifies both network structures and demographic predictors of sub-population membership, aiding personalised interventions. We evaluate our method through simulations and a youth mental health case study, demonstrating its potential to improve tailored interventions in health, education, and social policy.
Optimal Particle-based Approximation of Discrete Distributions
SEPTEMBER 2024
Particle-based methods include a variety of techniques, such as Markov Chain Monte Carlo (MCMC) and Sequential Monte Carlo (SMC), for approximating a probabilistic target distribution with a set of weighted particles. In this paper, we prove that for any set of particles, there is a unique weighting mechanism that minimizes the Kullback-Leibler (KL) divergence of the (particle-based) approximation from the target distribution, when that distribution is discrete - any other weighting mechanism (eg MCMC weighting that is based on particles’ repetitions in the Markov chain) is sub-optimal with respect to this divergence measure.
arXiv preprint: Optimal Particle-based Approximation of Discrete Distributions
Artificial and human intelligence for scientific discovery
JUNE 2024
This paper will discuss how we might develop AI systems which, together with our human brain, could transform scientific discovery. In order to do this, we need a definition of AI. AI is defined to be that field or industry which is at the intersection of data, algorithms, embedded in an application for the purpose of assisting decision-making.
Plug-and-Play Model-Agnostic Counterfactual Policy Synthesis for Deep Reinforcement Learning-Based Recommendation
NOVEMBER 2023
Recent advances in recommender systems have proved the potential of reinforcement learning (RL) to handle the dynamic evolution processes between users and recommender systems. However, learning to train an optimal RL agent is generally impractical with commonly sparse user feedback data in the context of recommender systems. To circumvent this, we propose to learn a general model-agnostic counterfactual synthesis (MACS) policy for counterfactual user interaction data augmentation. The counterfactual synthesis policy aims to synthesize counterfactual states while preserving significant information in the original state relevant to the user’s interests, building upon two different training approaches we designed: learning with expert demonstrations and joint training.
IEEE Transactions on Neural Networks and Learning Systems: Plug-and-Play Model-Agnostic Counterfactual
Policy Synthesis for Deep Reinforcement Learning-Based Recommendation
Fixed-distance Hamiltonian Monte Carlo
DECEMBER 2022
We propose a variation of the Hamiltonian Monte Carlo sampling (HMC) where the equations of motion are simulated for a fixed traversed distance rather than the conventional fixed simulation time.
Advances in Neural Information Processing Systems: Fixed-distance Hamiltonian Monte Carlo
Structure learning for hybrid bayesian networks
JUNE 2022
Bayesian networks have been used as a mechanism to represent the joint distribution of multiple random variables in a flexible yet interpretable manner. One major challenge in learning the structure of a Bayesian network is how to model networks which include a mixture of continuous and discrete random variables, known as hybrid Bayesian networks. This paper overviews the literature on approaches to handle hybrid Bayesian networks. Typically one of two approaches is taken: either the data are considered to have a joint distribution which is designed for a mixture of discrete and continuous variables, or continuous random variables are discretized, resulting in discrete Bayesian networks. In this paper, we propose a strategy to model all random variables as Gaussian, referred to it as Run it As Gaussian (RAG). We demonstrate that RAG results in more reliable estimates of graph structures theoretically and by simulation studies, than converting continuous random variables to discrete. Both strategies are also implemented on a childhood obesity data set. The two different strategies give rise to significant differences in the optimal graph structures, with the results of the simulation study suggesting that our strategy is more reliable.
arXiv preprint: Structure learning for hybrid bayesian networks
Bayesian optimization with informative parametric models via sequential Monte Carlo
MARCH 2022
Bayesian optimization (BO) has been a successful approach to optimize expensive functions whose prior knowledge can be specified by means of a probabilistic model. Due to their expressiveness and tractable closed-form predictive distributions, Gaussian process (GP) surrogate models have been the default go-to choice when deriving BO frameworks. However, as nonparametric models, GPs offer very little in terms of interpretability and informative power when applied to model complex physical phenomena in scientific applications. In addition, the Gaussian assumption also limits the applicability of GPs to problems where the variables of interest may highly deviate from Gaussianity. In this article, we investigate an alternative modeling framework for BO which makes use of sequential Monte Carlo (SMC) to perform Bayesian inference with parametric models. We propose a BO algorithm to take advantage of SMC’s flexible posterior representations and provide methods to compensate for bias in the approximations and reduce particle degeneracy. Experimental results on simulated engineering applications in detecting water leaks and contaminant source localization are presented showing performance improvements over GP-based BO approaches.
AdaptSPEC-X: Covariate-Dependent Spectral Modeling of Multiple Nonstationary Time Series
JANUARY 2022
We present the AdaptSPEC-X method for the joint analysis of a panel of possibly nonstationary time series. The approach is Bayesian and uses a covariate-dependent infinite mixture model to incorporate multiple time series, with mixture components parameterized by a time-varying mean and log spectrum. The mixture components are based on AdaptSPEC, a nonparametric model which adaptively divides the time series into an unknown number of segments and estimates the local log spectra by smoothing splines. AdaptSPEC-X extends AdaptSPEC in three ways. First, through the infinite mixture, it applies to multiple time series linked by covariates. Second, it can handle missing values, a common feature of time series which can cause difficulties for nonparametric spectral methods. Third, it allows for a time-varying mean. Through these extensions, AdaptSPEC-X can estimate time-varying means and spectra at observed and unobserved covariate values, allowing for predictive inference. Estimation is performed by Markov chain Monte Carlo (MCMC) methods, combining data augmentation, reversible jump, and Riemann manifold Hamiltonian Monte Carlo techniques. We evaluate the methodology using simulated data, and describe applications to Australian rainfall data and measles incidence in the United States.
HEALTH
Uncertainty and Inconsistency of COVID-19 Non-Pharmaceutical Intervention Effects with Multiple Competitive Statistical Models
JANUARY 2025
Quantifying the effect of non-pharmaceutical interventions (NPIs) is essential for formulating lessons from the COVID-19 pandemic. To enable a more reliable and rigorous evaluation of NPIs based on time series data, we reanalyse the data for the original official evaluation of NPIs in Germany using an ensemble of 9 competitive statistical methods for estimating the effects of NPIs and other determinants of disease spread on the effective reproduction number ℛ(t) and the associated error bars.
A causal artificial intelligence recommendation system for digital mental health
SEPTEMBER 2024
Digital mental health tools can improve access to care by offering interventional recommendations, often through rule-based systems or predictive AI. We developed CAIRS, a causal AI system that provides personalised recommendations by identifying intervention targets with the greatest impact on future outcomes.
arXiv preprint: A causal artificial intelligence recommendation system for digital mental health
A prognostic model for predicting functional impairment in youth mental health services
JULY 2024
This research developed and validated a prognostic model for youth mental health services to predict functional impairment trajectories over a 3-month period. This model serves as a foundation for further tool development and demonstrates its potential to guide indicated prevention and early intervention for enhancing functional outcomes or preventing functional decline.
Dynamic learning of individual-level suicidal ideation trajectories to enhance mental health care
JUNE 2024
There has recently been an increase in ongoing patient-report routine outcome monitoring for individuals within clinical care, which has corresponded to increased longitudinal information about an individual. However, many models that are aimed at clinical practice have difficulty fully incorporating this information, part due to the difficulty in dealing with the irregularly time-spaced observations that are common in clinical data. Consequently, we built individual-level continuous-time trajectory models of suicidal ideation for a clinical population (N = 585) with data collected via a digital platform. We demonstrate how such models predict an individual’s level and variability of future suicide ideation, with implications for the frequency that individuals may need to be observed.
The temporal dependencies between social, emotional and physical health factors in young people receiving mental health care-a dynamic Bayesian network analysis
JUNE 2023
The needs of young people attending mental healthcare can be complex and span multiple domains (e.g., social, emotional and physical health factors). MethodsDynamic Bayesian networks were used to examine the relationship between mental health factors across multiple domains in a longitudinal cohort of 2663 young people accessing youth mental health services.
Bayesian network modelling to identify on-ramps to childhood obesity
MARCH 2023
Childhood obesity is largely a function of socio-economic status, which is manifest through numerous downstream factors. Parental high school levels entangle with socio-economic status, and hence, are on-ramp to childhood obesity. The strong and independent causal relationship between birth weight and childhood BMI suggests a biological link. Our study implies that interventions that improve the socio-economic status, including through increasing high school completion rates, may be effective in reducing childhood obesity prevalence.
BMC Medicine: Bayesian netowkr modelling to identify on-rams to childhood obesity
Forecasting for COVID-19 has failed
APRIL 2022
Epidemic forecasting has a dubious track-record, and its failures became more prominent with COVID-19. Poor data input, wrong modeling assumptions, high sensitivity of estimates, lack of incorporation of epidemiological features, poor past evidence on effects of available interventions, lack of transparency, errors, lack of determinacy, consideration of only one or a few dimensions of the problem at hand, lack of expertise in crucial disciplines, groupthink and bandwagon effects, and selective reporting are some of the causes of these failures. Nevertheless, epidemic forecasting is unlikely to be abandoned. Some (but not all) of these problems can be fixed. Careful modeling of predictive distributions rather than focusing on point estimates, considering multiple dimensions of impact, and continuously reappraising models based on their validated performance may help.
International Journal of Forecasting: Forecasting for COVID-19 has failed
Learning as we go: An examination of the statistical accuracy of COVID19 daily death count predictions
MAY 2020
This paper provides a formal evaluation of the predictive performance of a model (and its various updates) developed by the Institute for Health Metrics and Evaluation (IHME) for predicting daily deaths attributed to COVID19 for each state in the United States.
The differential impact of major life events on cognitive and affective wellbeing
APRIL 2020
Major life events affect our wellbeing. However the comparative impact of different events, which often co-occur, has not been systematically evaluated, or studies assumed that the impacts are equivalent in both amplitude and duration, that different wellbeing domains are equally affected, and that individuals exhibit hedonic adaptation. We evaluated the individual and conditional impact of eighteen major life-events, and compared their effects on affective and cognitive wellbeing in a large population-based cohort using fixed-effect regression models assessing within person change.
GENERAL AND POLICY PAPERS
Bayesian causal discovery for policy decision making
MARCH 2025
This paper demonstrates how learning the structure of a Bayesian network, often used to predict and represent causal pathways, can be used to inform policy decision-making.
Data & Policy: Bayesian causal discovery for policy decision making
Panel stacking is a threat to consensus statement validity
SEPTEMBER 2024
Consensus statements without systematic evidence may be biased toward specific views. We describe this problem both generically and in detail, by a case study of a recent high-impact consensus statement about COVID-19.
Journal of Clinical Epidemiology: Panel stacking is a threat to consensus statement validity
Bayesian Adaptive Trials for Social Policy
JUNE 2024
This paper proposes Bayesian Adaptive Trials (BATs) as both an efficient method to conduct trials and a unifying framework for social policy interventions, addressing limitations inherent in traditional methods like Randomised Controlled Trials. BATs, grounded in decision theory, offer a dynamic "learning as we go" approach, enabiling the integration of diverse information types and facilitating a continuous, iterative process of policy evaluation.
Artificial and human intelligence for scientific discovery
JUNE 2024
This paper will discuss how we might develop AI systems which, together with our human brain, could transform scientific discovery. In order to do this, we need a definition of AI. AI is defined to be that field or industry which is at the intersection of data, algorithms, embedded in an application for the purpose of assisting decision-making.
Royal Society of NSW: Artificial and human intelligence for scientific discovery
Towards Risk-Free Trustworthy AI: Significance and Requirements
OCTOBER 2023
This review addresses key questions: why trustworthy AI is needed, the core requirements (e.g., explainability, fairness, privacy, and human oversight), and how trustworthy data can be ensured. It also examines trust priorities in complex applications, with examples from elds like IoT, robotics, and ntech. Practical insights are provided to help researchers build safer, more reliable AI systems that promote societal good.
How should a robot explore the Moon? A simple question shows the limits of current AI systems
JUNE 2023
AI systems should help humans make better, more accurate decisions. Yet even the most impressive and flexible of today’s AI tools – such as the large language models behind the likes of ChatGPT – can have the opposite effect.
Uncertainty: Nothing is more uncertain
APRIL 2022
This articles discusses the importance and types of uncertainty in environmental science. Uncertainty is categorized into four types, three of which are familiar to statisticians and one of which is introduced by the authors as knowledge uncertainty. The article claims that the Bayesian paradigm is a logically consistent mechanism, and a useful framework for decision makers in environmental science, to manage and quantify the first three types of uncertainty. However knowledge uncertainty, common in environmental sciences, requires more detailed thought and more nuanced management strategies for robust decision making in the environment. The article concludes with the observation that only if we acknowledge the uncertainty inherent in inferring complex quantities from data can we make robust and explainable policy decisions that build trust with the public.
ENVIRONMENT
Blockworlds 0.1.0: a demonstration of anti-aliased geophysics for probabilistic inversions of implicit and kinematic geological models
MAY 2022
Parametric geological models such as implicit or kinematic models provide low-dimensional, interpretable representations of 3-D geological structures. Combining these models with geophysical data in a probabilistic joint inversion framework provides an opportunity to directly quantify uncertainty in geological interpretations. For best results, care must be taken with the intermediate step of rendering parametric geology in a finite-resolution discrete basis for the geophysical calculation. Calculating geophysics from naively voxelized geology, as exported from commonly used geological modeling tools, can produce a poor approximation to the true likelihood, degrading posterior inference for structural parameters. We develop a simple integrated Bayesian inversion code, called Blockworlds, showcasing a numerical scheme to calculate anti-aliased rock properties over regular meshes for use with gravity and magnetic sensors. We use Blockworlds to demonstrate anti-aliasing in the context of an implicit model with kinematic action for simple tectonic histories, showing its impact on the structure of the likelihood for gravity anomaly.
Precipitation reconstruction from climate-sensitive lithologies using Bayesian machine learning
MAY 2021
This article presents an approach that links sedimentary deposits to reconstructed paleo-elevation maps. The authors use Bayesian machine learning to model the joint distribution of climate-sensitive sediments and annual precipitation through geological time. Their approach provides a set of 13 data-driven global paleo-precipitation maps between 14 and 249 Ma and captures major changes in paleo-precipitation as a function of plate tectonics, paleo-elevation and climate change.
Bayesreef: A Bayesian inference framework for modelling reef growth in response to environmental change and biological dynamics
MARCH 2020
Estimating the impact of environmental processes on vertical reef development in geological time is a very challenging task.
pyReef-Core is a deterministic carbonate stratigraphic forward model designed to simulate the key biological and environmental processes. This paper presents Bayesreef for the estimation and uncertainty quantification of parameters in pyReef-Core. The results show that Bayesreef accurately estimates and provides uncertainty quantification for parameters in pyReef-Core and establishes the groundwork for future research.
Multicore Parallel Tempering Bayeslands for Basin and Landscape Evolution
NOVEMBER 2019
In this paper, we extend Bayeslands using parallel tempering with high-performance computing to address previous limitations in Bayeslands. Our results show that parallel tempering Bayeslands not only reduces the computation time‚ but also provides an improvement in sampling multimodal posterior distributions, which motivates future application to continental scale landscape evolution models.
Bayeslands: A Bayesian inference approach for parameter uncertainty quantification in Badlands
NOVEMBER 2019
In this paper, the authors take a Bayesian approach for inference and uncertainty quantification of selected parameters in Badlands, which is a landscape evolution model simulates topography development at various space and time scales.Bayeslands fuses information from complex models with data, and prior knowledge. The results show that Bayeslands yields a promising distribution of the selected Badlands parameters.
Efficiency and robustness in Monte Carlo sampling for 3-D geophysical inversions with Obsidian v0.1.2: setting up for success
JULY 2019
Producing 3-D models of structures under the Earth's surface based on sensor data is a key problem in geophysics (for example, in mining exploration). There may be multiple models that explain the data well. We use the open-source Obsidian software to look at the efficiency of different methods for exploring the model space and attaching probabilities to models, leading to less biased results and a better idea of how sensor data interact with geological assumptions.
Climate inference on daily rainfall across the Australian continent, 1876-2015
OCTOBER 2018
The authors analyse over 294 million daily rainfall measurements since 1876, spanning 17,606 sites across continental Australia. They present examples of posterior inference on the mixture weights, monthly intensity levels, daily temporal dependence, offsite prediction of the effects of climate drivers and long-term rainfall trends across the entire continent.
