Discovery Early Career Researcher Award - Grant ID: DE240101190
Funder
Australian Research Council
Funding Amount
$451,000.00
Summary
Innovating and Validating Scalable Monte Carlo Methods. This project aims to develop innovative scalable Monte Carlo methods for statistical analysis in the presence of big data or complex mathematical models. Existing approaches to scalable Monte Carlo are only approximate, and their inaccuracies are difficult to quantify. This can have a detrimental impact on data-based decision making. The expected outcomes of this project are scalable Monte Carlo methods that are more accurate, fast and capa ....Innovating and Validating Scalable Monte Carlo Methods. This project aims to develop innovative scalable Monte Carlo methods for statistical analysis in the presence of big data or complex mathematical models. Existing approaches to scalable Monte Carlo are only approximate, and their inaccuracies are difficult to quantify. This can have a detrimental impact on data-based decision making. The expected outcomes of this project are scalable Monte Carlo methods that are more accurate, fast and capable of quantifying inaccuracies. Scientists and decision-makers will benefit from the ability to obtain timely, reliable insights for challenging applications.Read moreRead less
Scalable and Robust Bayesian Inference for Implicit Statistical Models. This project aims to develop the next generation of efficient methods for fitting complex simulation-based statistical models to data. Practitioners and scientists are interested in such implicit models to enable discoveries, produce accurate predictions and inform decisions under uncertainty. However, the associated computational cost has restricted researchers to implicit models that must have a small number of parameters ....Scalable and Robust Bayesian Inference for Implicit Statistical Models. This project aims to develop the next generation of efficient methods for fitting complex simulation-based statistical models to data. Practitioners and scientists are interested in such implicit models to enable discoveries, produce accurate predictions and inform decisions under uncertainty. However, the associated computational cost has restricted researchers to implicit models that must have a small number of parameters and be well specified, impeding scientific progress. This project will develop new computational methods and algorithms for implicit models that scale to high dimensions and are robust to misspecification. Benefits will arise from the more routine use of implicit models in epidemiology, biology, ecology and other fields.Read moreRead less
Advances in Sequential Monte Carlo Methods for Complex Bayesian Models. This project aims to develop efficient statistical algorithms for parameter estimation of complex stochastic models that currently cannot be handled. Parameter estimation is an essential component of mathematical modelling for answering scientific questions and revealing new insights. Current parameter estimation methods can be inefficient and require too much user intervention. This project will develop novel Bayesian alg ....Advances in Sequential Monte Carlo Methods for Complex Bayesian Models. This project aims to develop efficient statistical algorithms for parameter estimation of complex stochastic models that currently cannot be handled. Parameter estimation is an essential component of mathematical modelling for answering scientific questions and revealing new insights. Current parameter estimation methods can be inefficient and require too much user intervention. This project will develop novel Bayesian algorithms that are optimally automated and efficient by exploiting ever-improving parallel computing devices. The new methods will allow practitioners to process realistic models, enabling new scientific discoveries in a wide range of disciplines such as biology, ecology, agriculture, hydrology and finance.Read moreRead less
Statistical methods for quantifying variation in spatiotemporal areal data. This project aims to develop new statistical methods for extracting insights into spatial and temporal variation in areal data. These tools will extend the Australian Cancer Atlas which provides small area estimates for 20 cancers across Australia. The project is significant because it will allow government and other organisations to reap dividends from investment in collecting spatial information and it will enable mode ....Statistical methods for quantifying variation in spatiotemporal areal data. This project aims to develop new statistical methods for extracting insights into spatial and temporal variation in areal data. These tools will extend the Australian Cancer Atlas which provides small area estimates for 20 cancers across Australia. The project is significant because it will allow government and other organisations to reap dividends from investment in collecting spatial information and it will enable modelled small-area estimates to be released without compromising confidentiality. The expected outcomes include new statistical knowledge and new insights into cancer. The results will benefit the many disciplines, managers and policy makers that make decisions based on geographic data mapped over space and time. Read moreRead less
A Novel Approach to Semi-Supervised Statistical Machine Learning. Recent successes in the construction of classifiers for making diagnoses and predictions are due in part to their using much data labelled with respect to their class of origin. But typically there are little labelled data but plentiful unlabelled data. The goal of semi-supervised learning (SSL) is to leverage large amounts of unlabelled data to improve the performance using only small labelled datasets and so SSL is of paramount ....A Novel Approach to Semi-Supervised Statistical Machine Learning. Recent successes in the construction of classifiers for making diagnoses and predictions are due in part to their using much data labelled with respect to their class of origin. But typically there are little labelled data but plentiful unlabelled data. The goal of semi-supervised learning (SSL) is to leverage large amounts of unlabelled data to improve the performance using only small labelled datasets and so SSL is of paramount importance to applications where it is expensive or impractical to obtain much labelled data. The project is to develop a novel SSL approach that adopts a missingness mechanism for the missing labels to build a classifier that not only improves accuracy but it can be greater than if the missing labels were known.
Read moreRead less
Stochastic majorization--minimization algorithms for data science. The changing nature of acquisition and storage data has made the process of drawing inference infeasible with traditional statistical and machine learning methods. Modern data are often acquired in real time, in an incremental nature, and are often available in too large a volume to process on conventional machinery. The project proposes to study the family of stochastic majorisation-minimisation algorithms for computation of inf ....Stochastic majorization--minimization algorithms for data science. The changing nature of acquisition and storage data has made the process of drawing inference infeasible with traditional statistical and machine learning methods. Modern data are often acquired in real time, in an incremental nature, and are often available in too large a volume to process on conventional machinery. The project proposes to study the family of stochastic majorisation-minimisation algorithms for computation of inferential quantities in an incremental manner. The proposed stochastic algorithms encompass and extend upon a wide variety of current algorithmic frameworks for fitting statistical and machine learning models, and can be used to produce feasible and practical algorithms for complex models, both current and future.
Read moreRead less
Discovery Early Career Researcher Award - Grant ID: DE240100014
Funder
Australian Research Council
Funding Amount
$424,237.00
Summary
Causal relationship between taste and smell perception and eating behaviour. Around half of all Australians have a poor diet, which is a leading cause of many chronic conditions costing over $70 billion annually. This project aims to develop and apply novel statistical methods for determining the genetic basis of human taste and smell perception and its causal effects on eating behaviour. Expected outcomes include delivering new insights into such underlying individual differences for a wide ran ....Causal relationship between taste and smell perception and eating behaviour. Around half of all Australians have a poor diet, which is a leading cause of many chronic conditions costing over $70 billion annually. This project aims to develop and apply novel statistical methods for determining the genetic basis of human taste and smell perception and its causal effects on eating behaviour. Expected outcomes include delivering new insights into such underlying individual differences for a wide range of taste and olfactory traits; advanced analytical methods to assess causality; and a causal network of these sensory traits across over 100 consumable food items. From these outcomes, the benefits will be new strategies for improving food flavours and eating behaviours to enhance agri-food industry growth.Read moreRead less
Generalised Degrees of Freedom and Probabilistic Regularisation. This project intends to develop novel statistical tools for more accurate prediction by taking account of model complexity and uncertainties associated with the fitting procedure. The project also plans to develop a novel shrinkage approach via new penalty functions to avoid over-fitting and asymptotic properties. The key applications may include genetic studies where the number of predictors is large and biological experiments whe ....Generalised Degrees of Freedom and Probabilistic Regularisation. This project intends to develop novel statistical tools for more accurate prediction by taking account of model complexity and uncertainties associated with the fitting procedure. The project also plans to develop a novel shrinkage approach via new penalty functions to avoid over-fitting and asymptotic properties. The key applications may include genetic studies where the number of predictors is large and biological experiments where multivariate and temporal data are often collected – for example economical breeding in animal and fish farming and more effectively detecting the genes of interest in genetic studies on human, animals and plants.Read moreRead less