Asymptotic Geometric Analysis and Learning Theory. Learning Theory is used in various real-world applications in diverse research areas, ranging from Biology (e.g. DNA sequencing) to Information Sciences. Therefore, having a deep understanding of fundamental questions in Learning Theory, and in particular, pin-pointing the parameters that make a learning problem hard would have a significant practical impact. This projects aims to achieve this goal, and in addition, we expect it would have a hig ....Asymptotic Geometric Analysis and Learning Theory. Learning Theory is used in various real-world applications in diverse research areas, ranging from Biology (e.g. DNA sequencing) to Information Sciences. Therefore, having a deep understanding of fundamental questions in Learning Theory, and in particular, pin-pointing the parameters that make a learning problem hard would have a significant practical impact. This projects aims to achieve this goal, and in addition, we expect it would have a high theoretical value, as the questions we shall address are of independent interest to pure mathematicians.Read moreRead less
Asymptotic Geometric Analysis and Learning Theory. Learning Theory is used in various real-world applications in diverse research areas, ranging from Biology (e.g. DNA sequencing) to Information Sciences. Therefore, having a deep understanding of fundamental questions in Learning Theory, and in particular, pin-pointing the parameters that make a learning problem hard would have a significant practical impact. This projects aims to achieve this goal, and in addition, we expect it would have a hig ....Asymptotic Geometric Analysis and Learning Theory. Learning Theory is used in various real-world applications in diverse research areas, ranging from Biology (e.g. DNA sequencing) to Information Sciences. Therefore, having a deep understanding of fundamental questions in Learning Theory, and in particular, pin-pointing the parameters that make a learning problem hard would have a significant practical impact. This projects aims to achieve this goal, and in addition, we expect it would have a high theoretical value, as the questions we shall address are of independent interest to pure mathematicians.
Read moreRead less
Geometric parameters in Learning Theory. We aim to investigate the behaviour of geometric parameters which appear naturally in Statistical Learning Theory. Those parameters are used to control the sample complexity, which is the size of a random sample needed to produce an accurate prediction. They are also of independent interest in the local theory of Banach spaces. We shall use geometric methods originating in the local theory of Banach spaces to investigate the parameters and the way they in ....Geometric parameters in Learning Theory. We aim to investigate the behaviour of geometric parameters which appear naturally in Statistical Learning Theory. Those parameters are used to control the sample complexity, which is the size of a random sample needed to produce an accurate prediction. They are also of independent interest in the local theory of Banach spaces. We shall use geometric methods originating in the local theory of Banach spaces to investigate the parameters and the way they influence sample complexity. All the problems we focus on are not only important from the Machine Learning point of view, but are intriguing in their theoretical implications.Read moreRead less
Using data mining methods to remove uncertainties in sensor data streams. This project will develop key techniques for removing uncertainties in sensor data streams and thus improve the monitoring quality of sensor networks. The expected outcomes will benefit Australia by enabling improved, lower-cost monitoring of natural resources and management of stock raising.
Reconceiving Machine Learning. The proposed research will develop a new way to consider problems to which machine learning can be applied. Machine learning is crucial enabler of the digital economy. The research will provide better opportunities for Australian industry to gain a competitive advantage with machine learning technology. The framework developed will enable better opportunities for collaborative research and will build and strengthen international linkages.
High-performance computational data-mining techniques for feature detection in complex time series from large-scale, networked plasma experiments. Terabytes of data are gathered from large experimental facilities as complex time-series. Analysis of these data is daunting, especially when they involve high-dimensional spectral or image arrays. We will develop high-performance computational techniques for dimension reduction, efficient data-mining, and experimental control, using as an initial ta ....High-performance computational data-mining techniques for feature detection in complex time series from large-scale, networked plasma experiments. Terabytes of data are gathered from large experimental facilities as complex time-series. Analysis of these data is daunting, especially when they involve high-dimensional spectral or image arrays. We will develop high-performance computational techniques for dimension reduction, efficient data-mining, and experimental control, using as an initial target the H-1NF plasma fusion MNRF at the ANU and its >100 GB/year data stream. The techniques will immediately provide Australian researchers with unique tools for collaboration in international research to develop fusion as a low-emissions source of electricity, and will be applicable to complex time-series analysis in other areas of science, medicine, and defence.Read moreRead less
Network Intrusion Detection via Machine Learning. Computer security is an increasingly important, yet complex task. It
takes significant skills to configure systems properly such that they
are safe from malicious attacks.
The proposed project aims at designing automatic systems which are
able to adapt to an existing network configuration and which detect
novel and unusual events. For this purpose we will use modern machine
learning techniques, mainly based on kernels. In particular, rec ....Network Intrusion Detection via Machine Learning. Computer security is an increasingly important, yet complex task. It
takes significant skills to configure systems properly such that they
are safe from malicious attacks.
The proposed project aims at designing automatic systems which are
able to adapt to an existing network configuration and which detect
novel and unusual events. For this purpose we will use modern machine
learning techniques, mainly based on kernels. In particular, recently
developed algorithms to estimate the support of a distribution and
detect rare events will be employed in this context.
The project is in cooperation with Dr. Ralf Herbrich (Microsoft
Research, Cambridge).
Read moreRead less
Pattern Recognition and Scene Analysis via Machine Learning. We plan to use kernel methods, a novel machine learning technique, for computer vision problems, such as scene analysis and real time object recognition. Such capabilities are relevant for the design of intelligent and adaptive systems, suitable for complex real world environments. Expected outcomes are the design of efficient statistical tools which take the special nature of visual data into account (structure, decomposition, prior ....Pattern Recognition and Scene Analysis via Machine Learning. We plan to use kernel methods, a novel machine learning technique, for computer vision problems, such as scene analysis and real time object recognition. Such capabilities are relevant for the design of intelligent and adaptive systems, suitable for complex real world environments. Expected outcomes are the design of efficient statistical tools which take the special nature of visual data into account (structure, decomposition, prior knowledge of physical environments, etc.) and combine the advantages of feature based high-level vision methods with low-level machine learning techniques.
This proposal is part of a joint IST project with partners from the European Union.Read moreRead less
Analysing Iterative Machine Learning Algorithms with Information Geometric Methods. Online machine learning problems arise from situations where data is provided a point at a time. There are many classical algorithms for solving such problems based on the principle of stochastic gradient descent. Recent research by the CIs and others have thrown up interesting but diverse geometric connections that offer new insights. The proposed research aims to integrate the understanding of these algori ....Analysing Iterative Machine Learning Algorithms with Information Geometric Methods. Online machine learning problems arise from situations where data is provided a point at a time. There are many classical algorithms for solving such problems based on the principle of stochastic gradient descent. Recent research by the CIs and others have thrown up interesting but diverse geometric connections that offer new insights. The proposed research aims to integrate the understanding of these algorithms with the aim of designing algorithms better able to exploit prior knowledge, and to extend existing algorithms to new problem domains thus offering well principled and well understood algorithms for solving a variety of novel online problems.Read moreRead less
Finding and exploiting interesting paths in multidimensional information spaces. This project will invent a new approach for searching within a large complex information space, finding interesting paths between points within the space, visualising the results, and supporting rich, human-centric user interaction with queries and results. This project will embody these techniques in a novel, internet-scale framework to support rapid development of large path search and visualisation applications. ....Finding and exploiting interesting paths in multidimensional information spaces. This project will invent a new approach for searching within a large complex information space, finding interesting paths between points within the space, visualising the results, and supporting rich, human-centric user interaction with queries and results. This project will embody these techniques in a novel, internet-scale framework to support rapid development of large path search and visualisation applications. Evaluation will be via development of several exemplar applications. The techniques and framework will be applicable to a broad range of economically important problems in areas as diverse as health, travel, scientific publication search, product marketing and software engineering.Read moreRead less