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
Optimal Transforms of Random Vectors. This proposal focusses on development of optimal transforms to describe and model nonlinear phenomena when only statistical information is known. An optimal transform is a mathematical procedure that enables us to process information in a way that is most suited to the task in hand. These transforms have been successfully used in approximation, information theory, communications, control theory and signal and image processing. Applications include modelli ....Optimal Transforms of Random Vectors. This proposal focusses on development of optimal transforms to describe and model nonlinear phenomena when only statistical information is known. An optimal transform is a mathematical procedure that enables us to process information in a way that is most suited to the task in hand. These transforms have been successfully used in approximation, information theory, communications, control theory and signal and image processing. Applications include modelling of physical, chemical and biological systems, filtering and compression of signals and data classification and clustering. We propose two new hybrid models for realistic transforms in a general structural framework.
Read moreRead less
Information Geometry and Compressive Sensing for Radar and Communications. Australia's vast distances, thin population and extensive sea approaches force us to place heavy reliance on telecommunications and the remote sensing that radar and other modalities can provide. This project will enchance capabilities in sensing to provide more reliable, robust and cost effective communications and surveillance over a wide area.
A new perturbation method for solving singular operator equations with applications to complex systems. This project will develop new methods for analysis of web-based search routines such as Google PageRank, a new algorithm for optimal estimation of random signals, more accurate error analysis in the approximate solution of singular systems of equations and enhanced understanding of models for the simulated management of urban stormwater. The project will involve collaboration between two Aus ....A new perturbation method for solving singular operator equations with applications to complex systems. This project will develop new methods for analysis of web-based search routines such as Google PageRank, a new algorithm for optimal estimation of random signals, more accurate error analysis in the approximate solution of singular systems of equations and enhanced understanding of models for the simulated management of urban stormwater. The project will involve collaboration between two Australian universities and a leading European Research Institute. It will provide employment and vital training for two postdoctoral Research fellows and research projects for three postgraduate students and two honours students.Read moreRead less
A Computational Study of Nonconvex and Nonlinear Semi-infinite Optimisation Problems in Signal Processing. The operation of filtering is an important part of most modern communication engineering systems. Many important problems, which arise naturally from communications engineering applications, can be formulated as nonconvex optimization problems and nonlinear semi-infinite and/or semi-definite optimization problems. New optimization theory, in combination with novel computationally efficient ....A Computational Study of Nonconvex and Nonlinear Semi-infinite Optimisation Problems in Signal Processing. The operation of filtering is an important part of most modern communication engineering systems. Many important problems, which arise naturally from communications engineering applications, can be formulated as nonconvex optimization problems and nonlinear semi-infinite and/or semi-definite optimization problems. New optimization theory, in combination with novel computationally efficient solution methods, and efficient hardware implementation will be developed. The outcomes will enhance Australia's reputation in this cutting edge research and facilitate opportunity for international collaboration as well as commercial opportunity. The project will also provide an excellent environment for the training of junior researchers in the area.Read moreRead less
Robust Reformulation Methods. Many decision problems in engineering, business and economics are modeled as nonlinear continuous optimization problems. Often these are made difficult by the existence of constraints. In this project, we reformulate such problems as constrained nonsmooth equations, rather than optimization problems, and develop generalized Newton and quasi-Newton methods for solving them. The expected outcomes of this project include a systematic theory of reformulation methods, ....Robust Reformulation Methods. Many decision problems in engineering, business and economics are modeled as nonlinear continuous optimization problems. Often these are made difficult by the existence of constraints. In this project, we reformulate such problems as constrained nonsmooth equations, rather than optimization problems, and develop generalized Newton and quasi-Newton methods for solving them. The expected outcomes of this project include a systematic theory of reformulation methods, and robust and efficient algorithms for solving some important nonlinear continuous optimization problems. There is high potential for applications in engineering, business and finance.Read moreRead less
Quadratic Support Function Technique to Solving Hard Global Nonconvex Optimization Problems. Optimization techniques are becoming increasingly beneficial to modern Australian society in areas such as manufacturing and commerce by improving technical and management decisions. The proposed research is expected to produce enhanced optimization techniques that can be applied to solve a wider range of important problems too complex to be currently solved. The proposed research also represents an inte ....Quadratic Support Function Technique to Solving Hard Global Nonconvex Optimization Problems. Optimization techniques are becoming increasingly beneficial to modern Australian society in areas such as manufacturing and commerce by improving technical and management decisions. The proposed research is expected to produce enhanced optimization techniques that can be applied to solve a wider range of important problems too complex to be currently solved. The proposed research also represents an international collaboration which will improve Australia's ability to participate effectively in international research and innovation and to produce globally competitive mathematical technologiesRead moreRead less
Continuous Optimization with Linear Matrix Inequality Constraints. The proposed research is expected to lead to new insights and new joint collaborative work for both Autralian and Korean partners. Joining forces of the two teams will ensure that a full range of techniques can be utilized to provide rapid successful research outcomes. The proposed collaboration will give better opportunity to increase the visibility of the work from Korea in Australia, and vice versa. One of the key national be ....Continuous Optimization with Linear Matrix Inequality Constraints. The proposed research is expected to lead to new insights and new joint collaborative work for both Autralian and Korean partners. Joining forces of the two teams will ensure that a full range of techniques can be utilized to provide rapid successful research outcomes. The proposed collaboration will give better opportunity to increase the visibility of the work from Korea in Australia, and vice versa. One of the key national benefits is that the proposed research collaboration will provide extremly fertile ground for training postdoctoral researchers and graduate students in one of the most applicable areas of mathematics.Read moreRead less
Necessary and sufficient conditions for global minimum in multi-extremal global continuous optimization. A basic understanding of the mechanisms for finding local "best" (optimal) solutions has been
achieved through optimization techniques. However, solving global optimization problems, where we may have many local optimal solutions which are not the "absolutely best" (global), is vital for many applications in industry & science, and is intrinsically difficult. The lack of verifiable condition ....Necessary and sufficient conditions for global minimum in multi-extremal global continuous optimization. A basic understanding of the mechanisms for finding local "best" (optimal) solutions has been
achieved through optimization techniques. However, solving global optimization problems, where we may have many local optimal solutions which are not the "absolutely best" (global), is vital for many applications in industry & science, and is intrinsically difficult. The lack of verifiable conditions for a global optimum is a serious limitation. This project will develop verifiable such global optimality conditions for many classes of these problems. A new methodology, functional abstract convexity, developed by CIs and has shown promising results, will be extended and applied for solving these problems.Read moreRead less
A new improved solution to global optimization over multivariate polynomials: Mathematical principles, numerical methods and selected applications. Optimization technology is becoming increasingly beneficial to modern Australian society in areas such as wireless communications and manufacturing by improving performance or reducing costs. Our research will produce enhanced global optimization methodologies, capable of solving a wider range of problems that are currently too complex to be solved. ....A new improved solution to global optimization over multivariate polynomials: Mathematical principles, numerical methods and selected applications. Optimization technology is becoming increasingly beneficial to modern Australian society in areas such as wireless communications and manufacturing by improving performance or reducing costs. Our research will produce enhanced global optimization methodologies, capable of solving a wider range of problems that are currently too complex to be solved. Since global optimization technology is used in many scientific disciplines and modern industrial applications, the research will make many Australian science and industries more competitive. Our research also represents a program of high profile international collaborations that will improve Australia's ability to produce internationally competitive optimization technology.
Read moreRead less