Evolutionary multi-objective algorithms for Global Grids. This research investigates alternative software technologies for Grid-based evolutionary multi-objective decision algorithms. By employing the latest adaptive techniques and taking advantage of the low cost Grid infrastructure, new parallel evolutionary systems will be developed that can rapidly supply robust solutions to complex problems. This project will further develop an understanding of scaling issues in parallel evolutionary syste ....Evolutionary multi-objective algorithms for Global Grids. This research investigates alternative software technologies for Grid-based evolutionary multi-objective decision algorithms. By employing the latest adaptive techniques and taking advantage of the low cost Grid infrastructure, new parallel evolutionary systems will be developed that can rapidly supply robust solutions to complex problems. This project will further develop an understanding of scaling issues in parallel evolutionary systems and pave the way for even more widespread application of evolutionary techniques for large scale, data-intensive applications in science and industry.Read moreRead less
Unsupervised learning of finite mixture models in data mining applications. The extraction of useful information from massively large databases is known as data mining. Its broad but vague goal is to find "interesting structure" in the data, which typically leads to breaking the data into clusters. To this end, we consider the fast, efficient, and automatic learning of finite mixture models in hugh data sets without any prior knowledge of the structure. This probabilistic approach to the discove ....Unsupervised learning of finite mixture models in data mining applications. The extraction of useful information from massively large databases is known as data mining. Its broad but vague goal is to find "interesting structure" in the data, which typically leads to breaking the data into clusters. To this end, we consider the fast, efficient, and automatic learning of finite mixture models in hugh data sets without any prior knowledge of the structure. This probabilistic approach to the discovery and validation of group structure in data mining applications will considerably enhance knowledge management and decision support in science, industry, and government.
Read moreRead less
Realising the promise of neural networks for practical optimisation: improving their efficiency and effectivess through chaotic dynamics and hardware implementation. Combinatorial optimisation problems such as transportation routing and assembly-line scheduling are critical to the efficiency of many industries, but their combinatorial explosion makes rapid solution difficult. Neural networks (NNs) hold much potential for rapid solution though hardware implementation, but we need to improve the q ....Realising the promise of neural networks for practical optimisation: improving their efficiency and effectivess through chaotic dynamics and hardware implementation. Combinatorial optimisation problems such as transportation routing and assembly-line scheduling are critical to the efficiency of many industries, but their combinatorial explosion makes rapid solution difficult. Neural networks (NNs) hold much potential for rapid solution though hardware implementation, but we need to improve the quality of their solutions before developing hardware. We have previously shown that the rich dynamics of chaos can improve the efficiency and effectiveness of NNs. We aim to develop new chaotic NN models, rigorously evaluate them on industrially significant problems such as those arising in manufacturing, logistics and telecommunications, and demonstrate their speed through hardware acceleration.Read moreRead less
Parallel and Distributed Machine Learning - Smart Data Analysis in the Multicore Era. In large data centres our research will lead to reduced energy consumption by using graphics cards which have a much better computation to power ratio than traditional processors. On desktop computers, it will make machine learning practical by enabling efficient algorithms for spam filtering and content analysis. On networked systems it will lead to distributed inference, caching and collaborative filtering ap ....Parallel and Distributed Machine Learning - Smart Data Analysis in the Multicore Era. In large data centres our research will lead to reduced energy consumption by using graphics cards which have a much better computation to power ratio than traditional processors. On desktop computers, it will make machine learning practical by enabling efficient algorithms for spam filtering and content analysis. On networked systems it will lead to distributed inference, caching and collaborative filtering applications which will both reduced the bandwidth required and make the internet safer for users. Finally, it will enable rapid deployment of sensor networks for monitoring and detection, such as for environmental monitoring and safeguarding Australia's borders.Read moreRead less
3D Image segmentation and shape characterisation driven by topological persistence. Tomographic imaging is emerging as a new tool to help tackle a remarkable array of scientific challenges. What distinguishes healthy bone from that of osteoporosis sufferers? How does groundwater contamination spread? Why is a macadamia nut so hard to crack? What causes the iridescence in a butterfly wing? These are just a few of the questions being answered at tomographic facilities in Australia alone. By co ....3D Image segmentation and shape characterisation driven by topological persistence. Tomographic imaging is emerging as a new tool to help tackle a remarkable array of scientific challenges. What distinguishes healthy bone from that of osteoporosis sufferers? How does groundwater contamination spread? Why is a macadamia nut so hard to crack? What causes the iridescence in a butterfly wing? These are just a few of the questions being answered at tomographic facilities in Australia alone. By combining sophisticated mathematics with cutting edge image-processing algorithms, this project will yield a new class of topology driven image analysis techniques that will improve the accuracy and reliability of predictions made from tomographic images.Read moreRead less
Automated Determination of the Pose of a Human from Visual Information - Markerless 3D Pose Recovery of Humans from Videos. The development of 3D human pose recovery has been sought by computer vision researchers for many years. Our results will, firstly, have benefit for Australia's standing in the international computer vision community. Over time, the research outcomes will be developed into a software product for rehabilitation analysis by recognizing discrepancies between the walking pat ....Automated Determination of the Pose of a Human from Visual Information - Markerless 3D Pose Recovery of Humans from Videos. The development of 3D human pose recovery has been sought by computer vision researchers for many years. Our results will, firstly, have benefit for Australia's standing in the international computer vision community. Over time, the research outcomes will be developed into a software product for rehabilitation analysis by recognizing discrepancies between the walking patterns of healthy individuals and those with abnormalities as a result of accidents or diseases. The Australian economy will benefit by the reduction in the lifetime cost of injuries. This software will also provide benefits to the movie animation, computer games industry, and the training of athletes.Read moreRead less
Parameterized Algorithm Design and Complexity Analysis: New Methods and Strategic Applications in the FPT Algorithmic Server Project. A fundamental discovery of the first decades of computer science is that completely efficient (polynomial time) algorithms probably do not exist for thousands of natural computational problems. The project will result in new methods for designing and analyzing algorithms for hard problems with natural parameters, and in improved
algorithms for these problems.
Bayesian inference for complex regression models using mixtures. The project will use mixtures to flexibly model complex regression functions and will develop Bayesian methods for carrying out statistical inference on these models. The models will deal with both Gaussian and non-Gaussian data. Multiple explanatory variables are dealt with by mixing simple additives to produce flexible high dimensional function estimates. Variable selection and model averaging will be used to identify important v ....Bayesian inference for complex regression models using mixtures. The project will use mixtures to flexibly model complex regression functions and will develop Bayesian methods for carrying out statistical inference on these models. The models will deal with both Gaussian and non-Gaussian data. Multiple explanatory variables are dealt with by mixing simple additives to produce flexible high dimensional function estimates. Variable selection and model averaging will be used to identify important variables and thus make the estimation more efficient. The methods will be extended to multivariate responses where account will taken be taken of the structure of the dependence between responses.Read moreRead less
The Time-Varying Eigenvalue Problem with Application to Signal Processing and Control. Linear models are ubiquitous in representing physical processes. Decomposing a linear model into its fundamental components is known as the eigenvalue problem. In applications as wide ranging as astronomy, aircraft control systems, Internet search engines and communication systems, it is necessary to perform this decomposition of a pertinent time varying linear model on the fly. This project aims to develop si ....The Time-Varying Eigenvalue Problem with Application to Signal Processing and Control. Linear models are ubiquitous in representing physical processes. Decomposing a linear model into its fundamental components is known as the eigenvalue problem. In applications as wide ranging as astronomy, aircraft control systems, Internet search engines and communication systems, it is necessary to perform this decomposition of a pertinent time varying linear model on the fly. This project aims to develop significantly faster and more accurate algorithms for this time varying eigenvalue problem than currently exist. Very modern techniques will be employed to achieve this aim, and the potential benefits to Australian hi-tech industries are great.
Read moreRead less
A new theory for retinotectal map formation. How brains become wired up during development is a question of
importance to both biology and computing. In this project we adopt a
novel computational approach to understanding the development of
topographic maps, a wiring pattern that is ubiquitous in biological
nervous systems. This project will build capacity for research in
computational neuroscience in Australia. It may also lead to
technological benefits such as new ideas for the design o ....A new theory for retinotectal map formation. How brains become wired up during development is a question of
importance to both biology and computing. In this project we adopt a
novel computational approach to understanding the development of
topographic maps, a wiring pattern that is ubiquitous in biological
nervous systems. This project will build capacity for research in
computational neuroscience in Australia. It may also lead to
technological benefits such as new ideas for the design of self-wiring
computing devices, and new insights into
the causes of wiring defects both during normal development and
rewiring after injury.
Read moreRead less