Bayesian inference allows us to define a posterior distribution over the weights of a generic neural network (NN). Exact posteriors are usually intractable, in which case approximations can be employed. One such approximation - variational inference - is computationally efficient when using mini-batch stochastic gradient descent as subsets of the data are used for likelihood and gradient evaluations, though the approach relies on the selection of a variational distribution which sufficiently matches the form of the posterior. Particle-based methods such as Markov chain Monte Carlo and Sequential Monte Carlo (SMC) do not assume a parametric family for the posterior by typically require higher computational cost. These sampling methods typically use the full-batch of data for likelihood and gradient evaluations, which contributes to this computational expense. We explore several methods of gradually introducing more mini-batches of data (data annealing) into likelihood and gradient evaluations of an SMC sampler. We find that we can achieve up to $6\times$ faster training with minimal loss in accuracy on benchmark image classification problems using NNs.

Gaussian Process regression is a powerful non-parametric approach that facilitates probabilistic uncertainty quantification in machine learning. Distributed Gaussian Process (DGP) methods offer scalable solutions by dividing data among multiple GP models (or “experts”). DGPs have primarily been applied in contexts such as multi-agent systems, federated learning, Bayesian optimisation, and state estimation. However, existing research seldom addresses scenarios where the model inputs are uncertain — a situation that can arise in applications involving sensor noise or time-series modelling. Consequently, this paper investigates using a variant of DGP - a Generalised Product-of-Expert Gaussian Process - for the case where model inputs are uncertain. Three alternative approaches, and a theoretically optimal solution against which the approaches can be compared, are proposed. A simple simulated case study is then used to demonstrate that, in fact, neither approach can be guaranteed as optimal under all conditions. Therefore, the paper intends to provide a baseline and motivation for future work in applying DGP models to problems with uncertain inputs.

This work describes an Artificial Intelligence (AI)-based solution that predicts product quality when applied to a continuous manufacturing process. The proposed solution uses process parameters and product quality measurements that are obtained from a production line. The work detailed herein is problem-driven, showing an application within one of the UK’s foundation industries and identifying five key criteria an AI solution should ideally satisfy in continuous manufacturing applications; scalability, modularity, stable out-of-data performance, uncertainty quantification and robustness to unrepresentative data. The shortcomings, relative to these five criteria, of available AI approaches are discussed before a potential solution is presented. The proposed approach involves the application of a generalised product-of-expert Gaussian process whose noise model is constructed from a Dirichlet process. The ability of the model to fulfil the five key criteria and its performance when applied to the foundation industry case study is demonstrated.

Markov Chain Monte Carlo (MCMC) is a method for drawing samples from non-standard probability distributions. Hamiltonian Monte Carlo (HMC) is a popular variant of MCMC that uses gradient information to explore the target distribution. The Sequential Monte Carlo (SMC) sampler is an alternative sampling method which, unlike MCMC, can readily utilise parallel computing architectures. It is typical within SMC literature to target a tempered distribution using a proposal with an accept/reject mechanism. In this letter, we show how the proposal used in the No-U-Turn Sampler (NUTS), an advanced variant of HMC, can be incorporated into an SMC sampler without an accept/reject mechanism. Empirical results show that this can remove the need for tempering and gives rise to accurate estimates being generated in fewer iterations which motivates this technique being deployed on parallel hardware.

Manipulation of host plant physiologies by leaf-galling insects is a complex, multifaceted process. Among fundamental knowledge gaps surrounding this scientifically intriguing phenomenon is the appropriation of plant mineral nutrients and moisture for galling advantage. Small, soluble mineral ions and watery cell contents in dense gall tissues are easily dislocated or lost to routine sample preparation. In this study, an X-ray microanalysis was applied to investigate gall mineral nutrition. Morphologically diverse leaf galls were sampled from three Australian rainforest tree species. Using cryo-analytical scanning electron microscopy, real-time X-ray analytical maps of localized cellular mineral nutrients and water were integrated with anatomical images of gall and leaf cross-sectional surfaces to capture mineral-nutrient distribution patterns in situ. A comparison of host-leaf and gall anatomies bore direct evidence of drastic changes to leaf cells through the galling process. Distinct "wet" and "dry" regions within galls were anatomically and/or chemically differentiated, suggesting specific functionality. Wet regions comprising hydrated cells including soft gall-cavity linings where larvae are known to feed contained soluble mineral nutrients, while C-rich dry tissues largely devoid of mineral nutrients likely contribute structural support. The findings here provided otherwise inaccessible insights into leaf-gall mineral nutrition.

Antimicrobial resistance (AMR) emerges when disease-causing microorganisms develop the ability to withstand the effects of antimicrobial therapy. This phenomenon is often fueled by the human-to-human transmission of pathogens and the overuse of antibiotics. Over the past 50 years, increased computational power has facilitated the application of Bayesian inference algorithms. In this comprehensive review, the basic theory of Markov Chain Monte Carlo (MCMC) and Sequential Monte Carlo (SMC) methods are explained. These inference algorithms are instrumental in calibrating complex statistical models to the vast amounts of AMR-related data. Popular statistical models include hierarchical and mixture models as well as discrete and stochastic epidemiological compartmental and agent based models. Studies encompassed multi-drug resistance, economic implications of vaccines, and modeling AMR in vitro as well as within specific populations. We describe how combining these topics in a coherent framework can result in an effective antimicrobial stewardship. We also outline recent advancements in the methodology of Bayesian inference algorithms and provide insights into their prospective applicability for modeling AMR in the future.

Ozone (O-3) concentrations in the South Coast Air Basin (SoCAB) surrounding Los Angeles remain at unhealthy levels despite multiple decades of control programs designed to reduce emissions of precursor Volatile Organic Compounds (VOCs). Here we report on comprehensive VOC measurements made at Redlands, which has the highest measured O-3 concentrations in SoCAB, as part of the Re-Evaluating the Chemistry of Air Pollutants in California (RECAP-CA) field campaign (July-October 2021). Positive matrix factorization (PMF) analysis was applied to identify nine VOC factors. A photochemical chamber model initialized by field measurements was configured with a tagging technique to quantify the VOC factor contributions to O-3 formation in Redlands. Biogenic VOCs (BVOCs) made the largest contribution (26.6%) to O-3 formation, followed by traffic VOCs (21.2%), volatile chemical products (VCPs) (19%), and plant decomposition (14.9%). High O-3 episodes were not driven by increased VOC emissions from any single source, but rather were associated with stagnation events that concentrated VOCs from all sources and high temperature days that enhanced O-3 formation efficiency. This implies that VOC controls optimized to reduce O-3 concentrations would look similar in both the NOx-limited and VOC-limited regimes that can occur at Redlands. These results suggest that control strategies that reduce VOC and NOx emissions from the on-road vehicle fleet, such as increasing electrification, may yield O-3 reductions on days in both the NOx-limited and VOC-limited chemical regimes at Redlands.

With large wildfires becoming more frequent1,2, we must rapidly learn how megafires impact biodiversity to prioritize mitigation and improve policy. A key challenge is to discover how interactions among fire-regime components, drought and land tenure shape wildfire impacts. The globally unprecedented3,4 2019-2020 Australian megafires burnt more than 10 million hectares5, prompting major investment in biodiversity monitoring. Collated data include responses of more than 2,000 taxa, providing an unparalleled opportunity to quantify how megafires affect biodiversity. We reveal that the largest effects on plants and animals were in areas with frequent or recent past fires and within extensively burnt areas. Areas burnt at high severity, outside protected areas or under extreme drought also had larger effects. The effects included declines and increases after fire, with the largest responses in rainforests and by mammals. Our results implicate species interactions, dispersal and extent of in situ survival as mechanisms underlying fire responses. Building wildfire resilience into these ecosystems depends on reducing fire recurrence, including with rapid wildfire suppression in areas frequently burnt. Defending wet ecosystems, expanding protected areas and considering localized drought could also contribute. While these countermeasures can help mitigate the impacts of more frequent megafires, reversing anthropogenic climate change remains the urgent broad-scale solution.

Blood cultures are central to the management of patients with sepsis and bloodstream infection. Clinical decisions depend on the timely availability of laboratory information, which, in turn, depends on the optimal laboratory processing of specimens. Discrete event simulation (DES) offers insights into where optimization efforts can be targeted. Here, we generate a detailed process map of blood culture processing within a laboratory and use it to build a simulator. Direct observation of laboratory staff processing blood cultures was used to generate a flowchart of the blood culture laboratory pathway. Retrospective routinely collected data were combined with direct observations to generate probability distributions over the time taken for each event. These data were used to inform the DES model. A sensitivity analysis explored the impact of staff availability on turnaround times. A flowchart of the blood culture pathway was constructed, spanning labeling, incubation, organism identification, and antimicrobial susceptibility testing. Thirteen processes in earlier stages of the pathway, not otherwise captured by routinely collected data, were timed using direct observations. Observations revealed that specimen processing is predominantly batched. Another eight processes were timed using retrospective data. A simulator was built using DES. Sensitivity analysis revealed that specimen progression through the simulation was especially sensitive to laboratory technician availability. Gram stain reporting time was also sensitive to laboratory scientist availability. Our laboratory simulation model has wide-ranging applications for the optimization of laboratory processes and effective implementation of the changes required for faster and more accurate results. IMPORTANCE:Optimization of laboratory pathways and resource availability has a direct impact on the clinical management of patients with bloodstream infection. This research offers an insight into the laboratory processing of blood cultures at a system level and allows clinical microbiology laboratories to explore the impact of changes to processes and resources.

This paper reports on the initial implementation of Machine Learning (ML) for predicting the workload experienced by a pilot when performing a recovery to a naval ship. Pilots classify their workload for each landing by providing a subjective rating, which is used to determine the ship-helicopter operating limit (SHOL). Different workload metrics have been trialed to bridge the gap between pilot subjective ratings and objective flight data. With hundreds of different helicopter, ship and airwake parameters available to examine, ML provides an approach to understanding the complex interactions between these variables. This paper looks at the initial results obtained by applying ML techniques to train a classification algorithm with pilot control input data. Preliminary results showed 77.14% accuracy when training a Linear Discriminant algorithm to predict pilot workload from cyclic, collective, and pedal input data.

Biological volatilome analysis is inherently complex due to the considerable number of compounds (i.e., dimensions) and differences in peak areas by orders of magnitude, between and within compounds found within datasets. Traditional volatilome analysis relies on dimensionality reduction techniques which aid in the selection of compounds that are considered relevant to respective research questions prior to further analysis. Currently, compounds of interest are identified using either supervised or unsupervised statistical methods which assume the data residuals are normally distributed and exhibit linearity. However, biological data often violate the statistical assumptions of these models related to normality and the presence of multiple explanatory variables which are innate to biological samples. In an attempt to address deviations from normality, volatilome data can be log transformed. However, whether the effects of each assessed variable are additive or multiplicative should be considered prior to transformation, as this will impact the effect of each variable on the data. If assumptions of normality and variable effects are not investigated prior to dimensionality reduction, ineffective or erroneous compound dimensionality reduction can impact downstream analyses. It is the aim of this manuscript to assess the impact of single and multivariable statistical models with and without the log transformation to volatilome dimensionality reduction prior to any supervised or unsupervised classification analysis. As a proof of concept, Shingleback lizard (Tiliqua rugosa) volatilomes were collected across their species distribution and from captivity and were assessed. Shingleback volatilomes are suspected to be influenced by multiple explanatory variables related to habitat (Bioregion), sex, parasite presence, total body volume, and captive status. This work determined that the exclusion of relevant multiple explanatory variables from analysis overestimates the effect of Bioregion and the identification of significant compounds. The log transformation increased the number of compounds that were identified as significant, as did analyses that assumed that residuals were normally distributed. Among the methods considered in this work, the most conservative form of dimensionality reduction was achieved through analyzing untransformed data using Monte Carlo tests with multiple explanatory variables.

Environmental evaluations of metal nanoparticles (NP) rely on metal ion controls to distinguish between effects of the metal NP and its dissolution products. However, the coordinating or counter anion used in experimental controls may potentially influence biotic indicators used in ecotoxicology and soil health monitoring, compromising the ability to detect real nanoparticle effects and confounding interpretation of metal NP impacts. Using the example of copper oxide (CuO) NP, we demonstrate for the first time that depending on the anion used in the metal ion control (CuCl2 versus CuSO4), differing and even opposite conclusions may be drawn for CuO NP effects on a key microbiological indicator (enzyme activities) in environmental samples (soils). Moreover, this effect was specific to environmental conditions (soil management system) and indicator type (enzyme class), raising important methodological and interpretive implications for assessments of CuO NP impacts on soils. Our findings imply that assessments of soil health impacts of metal NP should consider multiple coordinating anion controls for a given metal, especially when the specific counterion is known to impact the biotic indicator (e.g., nutrient ions).

Lack of recent progress in reducing ground-level ozone (O3) concentrations to comply with health-based standards in the South Coast Air Basin (SoCAB) has motivated a reanalysis of emission control strategies. Here we used two parallel transportable smog chamber systems to measure the sensitivity of O3 to volatile organic compounds (VOCs) and nitrogen oxides (NOx = NO + NO2) in Pasadena and Redlands, California from July to October, 2021. The transportable smog chamber system measures the ambient O3 sensitivity and the ambient O3 chemical regime by comparing O3 formation in a basecase chamber and a perturbed chamber. The monthly median observed O3 sensitivity in Pasadena was stable in the VOC-limited regime, but showed a seasonal trend in Redlands, where median O3 sensitivity was VOC-limited in July and October and transitioned towards the NOx-limited regime in August and September. Day-specific O3 sensitivity at both Pasadena and Redlands could be either NOx-limited or VOC-limited on O3-nonattainment days. Calculated O3 isopleths for Pasadena and Redlands were constructed using a photochemical box model based on comprehensive measurements of NOx and VOCs during the Re-Evaluating the Chemistry of Air Pollutants in California (RECAP-CA) campaign. Calculated O3 isopleths were in good agreement with the chamber measurements. The calculations suggest that an additional ∼40% NOx reduction is needed for Pasadena and Redlands to move 95% of the days with O3 concentrations above 70 ppb to the NOx-limited regime where further NOx reductions will result in lower O3 concentrations.

BACKGROUND:Estimates of the prevalence of antimicrobial resistance (AMR) underpin effective antimicrobial stewardship, infection prevention and control, and optimal deployment of antimicrobial agents. Typically, the prevalence of AMR is determined from real-world antimicrobial susceptibility data that are time delimited, sparse, and often biased, potentially resulting in harmful and wasteful decision-making. Frequentist methods are resource intensive because they rely on large datasets. OBJECTIVES:To determine whether a Bayesian approach could present a more reliable and more resource-efficient way to estimate population prevalence of AMR than traditional frequentist methods. METHODS:Retrospectively collected, open-source, real-world pseudonymized healthcare data were used to develop a Bayesian approach for estimating the prevalence of AMR by combination with prior AMR information from a contextualized review of literature. Iterative random sampling and cross-validation were used to assess the predictive accuracy and potential resource efficiency of the Bayesian approach compared with a standard frequentist approach. RESULTS:Bayesian estimation of AMR prevalence made fewer extreme estimation errors than a frequentist estimation approach [n = 74 (6.4%) versus n = 136 (11.8%)] and required fewer observed antimicrobial susceptibility results per pathogen on average [mean = 28.8 (SD = 22.1) versus mean = 34.4 (SD = 30.1)] to avoid any extreme estimation errors in 50 iterations of the cross-validation. The Bayesian approach was maximally effective and efficient for drug-pathogen combinations where the actual prevalence of resistance was not close to 0% or 100%. CONCLUSIONS:Bayesian estimation of the prevalence of AMR could provide a simple, resource-efficient approach to better inform population infection management where uncertainty about AMR prevalence is high.

Recent studies in additive manufacturing (AM) monitoring techniques have focussed on the identification of defects using in situ monitoring sensor systems, with the aim of improving overall AM part quality. Much work has focussed on the use of of camera-based monitoring systems; however, limitations such as the slow response rates of the sensors (1-10kHz) and the post-processing requirements of the collected images make it difficult to apply these developmental monitoring methods on production systems in real-time. Furthermore, the replication of results from camera-based monitoring systems (often obtained using deep learning models) in a production environment is limited by the need for specialised hardware with high computational capacity (e.g GPUs). Focussing specifically on laser powder bed fusion ( PBF-L/M ), photodiodes, with fast data collection rates (50–100kHz) and providing data that is relatively easy to process are potentially better suited to real-time monitoring systems. The current study, therefore, focuses on using data collected from photodiodes to identify defects in PBF-L/M builds. A predictive model with real-time potential is proposed that, having been validated on data from computer tomography (CT) images, can be used to locate porosity within layers of PBF-L/M builds.

The factors that influence population structure and connectivity are unknown for most terrestrial invertebrates but are of particular interest both for understanding the impacts of disturbance and for determining accurate levels of biodiversity and local endemism. The main objective of this study was to determine the historical patterns of genetic differentiation and contemporary gene flow in the terrestrial snail, Austrochloritis kosciuszkoensis (Shea & O. L. Griffiths, 2010). Snails were collected in the Mt Buffalo and Alpine National Parks in Victoria, in a bid to understand how populations of this species are connected both within continuous habitat and between adjacent, yet separate environments. Utilising both mitochondrial DNA (mtDNA) and single nucleotide polymorphism (SNP) data, the degree of population structure was determined within and between sites. Very high levels of genetic divergence were found between the Mt Buffalo and Alpine snails, with no evidence for genetic exchange detected between the two regions, indicating speciation has possibly occurred between the two regions. Our analyses of the combined mtDNA and nDNA (generated from SNPs) data have revealed patterns of genetic diversity that are consistent with a history of long-term isolation and limited connectivity. This history may be related to past cycles of changes to the climate over hundreds of thousands of years, which have, in part, caused the fragmentation of Australian forests. Within both regions, extremely limited gene flow between separate populations suggests that these land snails have very limited dispersal capabilities across existing landscape barriers, especially at Mt Buffalo: here, populations only 5 km apart from each other are genetically differentiated. The distinct genetic divergences and clearly reduced dispersal ability detected in this data explain the likely existence of at least two previously unnamed cryptic Austrochloritis species within a 30-50 km radius, and highlight the need for more concentrated efforts to understand population structure and gene flow in terrestrial invertebrates.

Many ecosystems globally evolved with fire. However, there is a gap in our knowledge regarding the effect of fires on less mobile invertebrates. Land snails in Australia are a diverse group with a high level of endemism, but we understand very little of their ecology, especially how they are affected by large fires. We studied the short-term (one year after a fire event) response of 1) land snail species composition, 2) snail abundances, and 3) impacts on 18 priority species to an unprecedented large fire event throughout south-eastern Australia in 2019/2020 (Black Summer). Our study ranged over >100,000 km2, surveying162 sites. Our study revealed that land snail species composition changed significantly depending on fire severity class, regardless of habitat type (rainforest or eucalypt forest) or climate. Medium and high severity fire caused snail abundance to decline significantly, with micro-snails being more sensitive to medium and high severity fires. Abundance dropped by 75 % and 64 % for micro-, and macro-snails respectively, but low severity fire did not have a significant impact on abundance. For most priority species, fires affected more than half of their known extent of occurrence although most species were found in all fire severities. However, severity category was an important driver in determining the probability of occurrence of priority species. To conserve Australia's land snail species, we must know how they are affected by disturbances. Our results suggest that the total area and severity of fires will determine the impact of fires on land snails.

The 2019–2020 megafires in eastern Australia have devastated large parts of the known distributional range of the minute land snail Paralaoma annabelli, prompting conservation concerns for this species. However, this species is poorly defined thus hampering its accurate identification and the delineation of its distribution. Most crucially, it has been questionable if and how P. annabelli could be distinguished from another Australian congener, Paralaoma morti. This systematic ambiguity posed a problem in assessing the impact of the 2019–2020 wildfires in Australia on this species. Herein, we demonstrate, based on comparative morphometrics as well as analyses of mitochondrial and nuclear DNA, that P. annabelli is indeed distinct from a second widespread species of Paralaoma, which is identified as P. morti by some workers. Yet, sequences of P. morti cluster closely with non-Australian sequences of the globally distributed species P. servilis. Therefore, the taxonomic status of P. morti in relation to P. servilis remains to be investigated. Our comparative morphological analyses revealed that P. annabelli is significantly smaller than P. morti, has a significantly flatter shell, more elongated aperture, lower spire, and tighter coiling whorls. With the revised diagnosis of P. annabelli, we have delineated its distribution in New South Wales based on the examination of all available museum samples. We show that P. annabelli is primarily found at higher elevations in the Great Dividing Range while P. morti is widespread in eastern Australia. In addition, molecular phylogenetic analyses reveal that the genera Pseudiotula, Iotula, Trocholaoma and Miselaoma, all described based only on shell characteristics, form a single clade with the abovementioned species of Paralaoma. This reveals the inadequacies of a purely shell-based taxonomy in punctids and highlights the need for a more integrative approach to punctid systematics.

Using contingency theory, we argue that there is not a uniform approach for companies to govern information technology (IT) investments. Rather, the level of governance over IT investments is contingent upon the organization’s goals for its IT investments. We find that Australian organizations with both operation- and market-focused IT investment goals (i.e. dual-focused IT goals) demonstrate higher IT investment governance (ITIG) levels than those with less focused IT goals. We also document that dual-IT-focused firms that do not implement high levels of ITIG underperform. Our study informs business executives, boards of directors, and other practitioners interested in governance implementations over IT investments. JEL Classification: M1

AbstractIn many parts of the world, livestock production and biodiversity conservation are important land uses of native grasslands in agricultural landscapes. Approaches to managing grasslands typically differ between production farms and conservation areas as they have different goals. Such differences may have consequent effects on the spatial and temporal habitat suitability for grassland fauna. In semi‐arid grasslands of south‐eastern Australia, the critically endangered Plains‐wanderer (Pedionomus torquatus) is a grassland habitat‐specialist bird that can occur on land managed for livestock production and conservation, but it is unclear if, and when, habitat suitability is affected in each land‐use type.Here, we investigate how land‐use type (livestock production, conservation) and rainfall (preceding accumulated rainfall) affect habitat suitability for the Plains‐wanderer using 11 years of bird occurrence and remotely sensed habitat structure data.We found habitat suitability for the Plains‐wanderer was driven by an interaction between land use and rainfall, with conservation areas supporting larger areas of preferred habitat structure during dry periods but less during wet periods. By contrast, Plains‐wanderers were more likely to occur on livestock production farms during wet periods. We speculate this is because higher grazing pressure on livestock production farms was able to limit biomass accumulation and, hence, maintain more areas of preferred habitat structure.Our findings show that land used for livestock production can complement conservation areas by providing preferred habitat for the Plains‐wanderer during climatic periods that promote grass growth. Furthermore, we highlight that land use and climate are important temporal drivers of grassland dynamics, and approaches to biodiversity conservation should consider how patterns of habitat suitability may shift across landscapes over time. Strategic, landscape‐scale planning and effective agri‐environmental initiatives will be critical to the future of grassland birds such as the Plains‐wanderer.