2006 seminar series

Biography:

Owen completed his undergraduate degree in Computer Science at MonashUniversity and is currently conducting his PhD in the field of Evolutionary Ethics. Owen's PhD supervisors at Monash University are Dr Kevin Korb and Dr Ann Nicholson.

Abstract:

Gould and Eldridge's (1972) "Punctuated Equilibria: AnAlternative to Phyletic Gradualism" drew attention to what they saw as amistaken inference -- that because evolution can only occur gradually,evolution can only occur at a constant, continuous rate, a concept theylabelled Phyletic Gradualism. Gould and Eldridge argued instead that mostevolution occurs during short (geologically speaking) speciation events,with species exhibiting stasis for the vast majority of the time. ThePunctuated Equilibrium hypothesis describes a species selection mechanismsimilar to the group selection model of Wynne-Edwards (1962). In ourstudies, we have developed an ALife simulation environment where agents,with evolving mating signatures, interact with evolving food niches. Theresult is the formation, through peripatric speciation, of reproductivelyisolated sub populations (i.e. species), creating a selection mechanismwhich could potentially support the evolution of altruistic traits viagroup/species selection.

Biography:

Tim completed his undergraduate Comp Sci degree at Melbourne University and his PhD at the University of Sydney. Most of his research to date has involved new paradigms and algorithms for network visualisation.

Abstract:

Graph drawing is the field of computer science concerned with automatically generating visualisations of relational network data. It is an important tool for understanding networks such as: biological pathways; complex software designs; social networks such as crime organisations; or communications infrastructure. A popular family of algorithms for automatic graph layout attempt to arrange the graph by minimising a continuous energy function. We will discuss how mathematically sound constraint optimisation methods may be introduced into these algorithms. Compared to existing graph drawing methods this results in layout of higher quality that is more easily customised to specific applications.

Biography:

Ah Chung Tsoi was born in Hong Kong. He received the Higher Diploma in Electronic Engineering from Hong Kong Technical College in 1969, and the M.Sc. degree in electronic control engineering, and Ph.D. degree in control engineering from University of Salford in 1970 and 1972 respectively. He also received a B.D. degree from University of Otago in 1980.
From 1972 to 1974 he was a Senior Research Fellow at the Inter-University Institute of Engineering Control, University College of North Wales, Bangor, Wales. From 1974 to 1977 he was a Lecturer at the Paisley College of Technology, Paisley Refrewshire, Scotland. From 1977 to 1985 he was a Senior Lecturer at the University of Auckland, New Zealand. From 1985 to 1990, he was a Senior Lecturer at the University College, University of New South Wales. From 1990 to 1996, he was an Associate Professor, and then a Professor in Electrical Engineering at the University of Queensland, Australia. While at the University of Queensland, Professor Tsoi led a large R & D project on speaker verification. From July 1996 to Feb 2004 he was at the University of Wollongong, where he had been Dean, Faculty of Informatics (1996 to 2001), and Director of Information Technology Services (1999 to 2001). From February 2001 to 2004, he was the foundation Pro Vice Chancellor (Information Technology and Communications) at the University of Wollongong. In the role of Pro Vice-Chancellor (Information Technology and Communications) he initiated a number of infrastructure projects: gigabit ethernet for the campus, voice over IP telephony, wireless access for the entire campus, implementation of a student management package. His research interests include aspects of neural networks and fuzzy systems and their application to practical problems, adaptive signal processing, speech processing, adaptive control. He applies these techniques to many practical application situations, including data mining, internet search engine designs. From Feb 2004 to Dec 2005, he was the Executive Director, Mathematics, Information and Communications Sciences, Australian Research Council (ARC). In this role, he oversaw grant administration as well as formulation of policies pertaining to Mathematics, Information and Communication Sciences research in Australia. He also initiated a funding scheme, e-Research Support under the Special Research Initiatives to support e-Research activities in Australia. Since Dec 2005, he has been the Director, e-Research Centre, at Monash University.
He publishes widely in the area of adaptive signal processing, neural networks, fuzzy systems. He received an Outstanding Alumni Award, Hong Kong Polytechnic University in 2001.

Abstract:

In this talk, we will review some of the difficulties in clustering an XML document corpus. XML documents are semi-structured. There are a number of possibilities in clustering the documents: (a) structure information alone, (b) content information alone, and (c) structure and content information. We will introduce two methods, one an unsupervised learning method called self organising map with structured data (SOM-SD), and the other one a supervised learning method called graphic neural network (GNN). We will show that it is possible to cluster the document corpus using both methods, with structure information only. However we are less successful in using both structure and content information.

Biography:

David Albrecht is currently a Senior Lecturer in the Clayton School of IT. He received his BSc. and Ph.D. degrees in Mathematics from Monash University. He has worked in a broad range of areas, including User Modelling, Machine Learning, Bayesian Networks, Statistics, Logic, Functional Analysis, and Optimization. His current interest are primary in Bayesian Statistics and Optimization.

Abstract:

The limitations of traditional knowledge representation methods for modelling human behaviour has led to the investigation of statistical models. Predictive statistical models enable the anticipation of certain aspects of human behaviour, such as goals, actions and preferences. In this talk I will give an overview of the various statistical techniques used in user modelling, and review some of the main outstanding issues.

Download the presentation (ppt, 552kb)

Biography:

Jon McCormack is an Australian-based electronic media artist and researcher in Artificial Life and Evolutionary Music and Art. His research interests include generative evolutionary systems, machine learning, L-systems and developmental models. McCormack is also a practising electronic media artist. He holds an Honours degree in Applied Mathematics and Computer Science from Monash University, a Graduate Diploma of Art from Swinburne University and a PhD in Computer Science from Monash University. He is currently Senior Lecturer in Computer Science and co-director of the Centre for Electronic Media Art (CEMA) at Monash University in Melbourne, Australia. CEMA is an interdisciplinary research centre established to explore new collaborative relationships between computing and the arts.
His artworks have been exhibited internationally a wide variety of galleries, museums and symposia, including the Museum of Modern Art (New York, USA), Tate Gallery(Liverpool, UK), ACM SIGGRAPH (USA), Prix Ars Electronica (Austria) and the Australian Centre for the Moving Image (Australia). His work has received a number of international awards for new media art including prizes at Ars Electronica (Austria), Images du Futur (Canada), New Voices, New Visions(USA), Alias/Wavefront(USA) and Nagoya Biennial (Japan). The monograph Impossible Nature: the art of Jon McCormack was published by the Australian Centre for the Moving Image in 2005, and documents McCormack’s creative achievements over the last 15 years.

Abstract:

In this talk I will give an overview of the evolutionary systems I have used over the last ten years to 'grow' and generate complex organic form. L-systems are parallel rewriting grammars, originally developed in the 1960s to model the cellular development of algae. In the last thirty years a number of important developments have taken place, giving modern interpretations of L-systems significant success in modelling a variety of complex organic and structural phenomena. I will explain a number of these developments, including my own Cellular Developmental Model: a generalised system for simulating growth and development. I will also briefly introduce a novel interactive, artificial life ecosystem, where artificial creatures use Learning Classifier Systems (LCS) to evolve symbiotic relationships with a human audience.
I will conclude with some ideas on creative intelligence, particularly in relation to issues of embodiment, functional morphology and morphological computation.

Download the presentation (pdf, 3.85mb)

Biography:

Dr Ying Yang obtained PhD in computer science in 2003 from Monash University, Australia. She then worked as a research fellow in the department of computer science, University of Vermont, USA. Currently she holds a position of research fellow in the Clayton school of IT, Monash University, Australia. Her research focuses on data mining and machine learning.
http://www.csse.monash.edu.au/~yyang/publications.html

Abstract:

In many online applications of machine learning, the computational resources available for classification will vary from time to time. Most techniques are designed to operate within the constraints of the minimum expected resources and fail to utilize further resources when they are available. We propose a novel anytime classification approach, which is capable of delivering strong prediction accuracy with little CPU time and utilizing additional CPU time to increase classification accuracy.

Download the presentation (pdf, 1.4mb)

Biography:

Dr Adam Kowalczyk has obtained MSc in applied mathematics in 1975 and PhD in mathematics in 1978, both from Warsaw University of Technology. He worked in as lecturer in mathematics for Warsaw University Technology (1978-1981) and University of Baghdad, Iraq (1981-1983) before migrating to Australia. Here he worked for Telstra Research Laboratories in 1984-2003 and Peter Mac Callum Cancer Centre 2003-04, both in Melbourne, Australia. Currently he holds a position of Principal Researcher with Life Sciences, NICTA, Victoria Research Laboratories, establishing bio-informatics research group.

His research focus evolved over years from differential geometry through singularities of function and their applications in physics to machine learning and its applications to telecommunications, biology and medicine.

Adam Kowalczyk has earned a number of distinctions, including Distinguished Research Fellow in 1999 from Telstra and the winner of KDD Cup 2002, the prestigious international data mining competition, for co-development of the most accurate predictor for Yeast Gene Regulation Prediction task.

Abstract:

One of the central objectives of supervised analysis of microarray data is the generation of compact and accurate models that facilitate the design of follow up wet lab experiments. Typically this is achieved by aggressive feature pre-selection, such as the t-test or signal-to-noise filtering, prior to application of a classification algorithm such as k-nearest neighbors or the support vector machine. We show a number of real life and synthetic examples where such routine data mining procedures do not provided expected results. Moreover, we demonstrate that the observed accuracy of the final predictors can be systematically worse than the random guessing. Some natural examples of classification of cancer survival genomic datasets as well as synthetic examples of such datasets will be presented. These counter-intuitive prediction outcomes will be linked to the specific geometrical configuration of the datasets. Finally, we also discuss a number of counter-intuitive remedies which we have discovered while working with some of these datasets.

Download the presentation (ppt, 2.41mb)

Biography:

After receiving his PhD from the University of Sydney, Australia, Dr. Ting had worked at the University of Waikato, New Zealand and Deakin University, Australia. He joins Monash University since 2001. He is now an Associate Professor in Gippsland School of Information Technology. His research interests include Machine Learning (cost-sensitive learning, model combination, classifier evaluation), Data Mining, Content-Based Image Retrieval and Support Vector Machines.

Abstract:

This talk reports a recent work on decision tree ensembles that employ randomisation to generate multiple models, in particular decision trees that are generated using a complete-random process. This work is motivated to find out the reasons why complete-random tree ensembles work in practice. It begins with an investigation on the strengths and weaknesses of complete-random tree ensembles. That leads us to introduce a variable randomness that allows the ensembles to work well in different kinds of data. Experimental results show that the proposed ensemble approach is significantly better than existing approaches such as Random Forests and Max-diverse Ensemble, and it is comparable to the state-of-the-art C5 boosting.

Download the presentation (pdf, 1.77mb)

Biography:

Dr. Nathalie Japkowicz is an Associate professor of Computer Science in the School of Information Technology and Engineering at the University of Ottawa. She is a visiting professor at Monash University, Clayton during the 2006-2007 Canadian School Year. She obtained her Ph.D. from Rutgers University (New Jersey) in 1999. Her area of study is Machine Learning with special emphases on the class imbalance problem, one-class learning, machine learning applied to computer and nuclear security, text mining, and, more recently, performance evaluation for machine learning.

Abstract:

The evaluation of classifiers or learning algorithms is not a topic that has, generally, been given much thought in the fields of Machine Learning and Data Mining. More often than not, common off-the-shelf metrics such as Accuracy, Precision/Recall and ROC Analysis as well as confidence estimation methods, such as the t-test, are applied without much attention being paid to their meaning. The purpose of this talk is to underline some of the problems that can arise from our current practices and suggest that it might be useful, in certain cases, to either borrow evaluation methods from other fields or create our own measures. We will specifically look at two different types of domains, domains in which the two classes are as important, and domains suffering from a severe class imbalance and show how some of the methods stemming from the field of biostatistics can be of use to us in these cases. We will conclude the talk by presenting a novel evaluation method designed for the severe class imbalance case.

Download the presentation (ppt, 975kb)

Biography:

Geoff Webb holds a research chair in the Faculty of Information Technology at Monash University. Prior to Monash he held appointments at Griffith University and then Deakin University where he received a personal chair. His primary research areas are machine learning, data mining, and user modelling. He is widely known for his contribution to the debate about the application of Occam's razor in machine learning and for the development of numerous algorithms and techniques for machine learning, data mining and user modelling. His commercial data mining software, Magnum Opus, is marketed internationally by Rulequest Research. He is editor-in-chief of Data Mining and Knowledge Discovery and a member of the editorial boards of Machine Learning, ACM Transactions on Knowledge Discovery in Data, User Modeling and User-Adapted Interaction, and Knowledge and Information Systems.

Abstract:

Naive Bayes is an extremely efficient classification learning technique. Despite its simplicity, naive Bayes has proved remarkably accurate for many tasks. In consequence, it is widely deployed, even though its accuracy is known to be limited by its attribute independence assumption. Of numerous proposals to improve the accuracy of naive Bayes by weakening its attribute independence assumption, both LBR and SuperParent TAN have demonstrated remarkable accuracy. However, both techniques attain this accuracy at a considerable computational cost. Motivated by both theoretical and practical considerations, we present a new approach to weakening the attribute independence assumption by averaging all of a constrained class of semi-naive Bayesian classifiers. In extensive experiments this technique delivers comparable prediction accuracy to LBR and SuperParent TAN, with substantially improved computational efficiency. It has the desirable properties of

1. - training time that is linear with respect to training set size,
2. - supporting parallel and anytime classification, and
3. - allowing incremental learning.

Despite being generative, it delivers classification accuracy competitive with state-of-the-art discriminative techniques.

Download the presentation (pdf, 279kb)

Biography:

Hari Koesmarno is a data miner and business modeller at the Australian Taxation Office. He completed his B.Sc at the University of Canterbury in 1989 and Master (Hons) at Lincoln University in 1991. He has been involved in research projects with the New Zealand Forest Research Institute/University of Canterbury, New Zealand and the Fujitzu parallel computing project with Australian National University. Since 1998 Hari has been working with Australian Government in the area of statistics, modelling and data mining . He has been published in over twenty international publications/journals including in the Journal of Applied Statistics, Mathematical and Computer Modelling, Environmental Software and Journal of Agricultural Science, Cambridge.

Abstract:

Techniques currently used and techniques which have several potential applications for ATO business management are discussed. The latter is proposed for Ph.D research at Monash. The ATO applications discussed are mainly on (a) identifying client risk in reporting and (b) management of ATO resources for intervention of clients with risk in lodgement. The methodology being developed for data mining (include pre and post data mining) are (i) modelling frame work; (ii) stratification/segmentation of income/tax distribution for risk on reporting; (iii) improving strike rate using ensemble model; (iv) methodology for resource management for lodgement compliance; and (v) dynamic improvement on risk score using spline, moving average and bootstrap.

Biography:

Ann Nicholson is a Senior Lecturer in the School of Computer Science and Software Engineering at Monash University. She received her B.Sc

(Hons) and M.Sc. degrees in Computer Science from the University of Melbourne. In 1992 she received her Ph.D. in Engineering from the University of Oxford, where she was part of the Robotics Research Group. After 2 years as a post-doctoral research fellow in Computer Science at Brown University, she took up a position at Monash University in 1994. Her areas of research interest are reasoning under uncertainty, Bayesian networks, knowledge engineering, user modelling and plan recognition, stochastic planning and monitoring, and intelligent agents.

Abstract:

Bayesian networks (BNs) have become a popular AI representation for reasoning under uncertainty, with successful applications in a range of domains. Most successful applications to date have been built through knowledge elicitation from experts, however this is difficult and time consuming, which has lead to the recent interest in automated methods for learning BNs from data. There is as yet no established methodology for a hybrid approach which combines knowledge elicitation from experts and automated knowledge discovery methods.

In this seminar, I will begin with a brief introduction to Bayesian networks and illustrate their workings in a commericially available software package. I will then give an overview of the BN applications developed here at Monash using a hybrid approach by myself and Kevin Korb These applications include intelligent tutoring, Bayesian poker, weather forecasting, ecological risk assessment and medical risk assessment.

Download the presentation (ppt, 1.36mb)

Biography:

Richard Brent was appointed Foundation Professor of Computer Science at the Australian National University in 1978. In March 1998 he became the Statutory Professor of Computing Science in the University of Oxford. In March 2005 he took up a 5-year position as ARC Federation Fellow in the Mathematical Sciences Institute and the Research School of Information Sciences and Engineering at the Australian National University.

http://wwwmaths.anu.edu.au/~brent/qual.html

Abstract:

I will describe some algorithms for testing primality, and show why a randomised algorithm that sometimes gives the wrong answer is preferable in practice to a deterministic polynomial-time algorithm that always gives the correct answer.

Download the presentation (pdf, 147kb)

Biography:

Andrew P. Paplinski received the M.Eng and Ph.D. degrees from the Faculty of Electronics, Warsaw University of Technology, Poland, in 1967 and 1980, respectively.He is currently Associate Professor in the Monash University, Melbourne, Australia.

His recent research activities concentrate around computational neuroscience, computational intelligence and computer vision. The other areas of expertise include signal and image processing, ultrasonic imaging, neural networks and aspects of hardware implementation of the related algorithms.He published over ninety papers, book chapters and technical reports and five textbooks.

Currently he collaborates with Prof. Lennart Gustafsson from the Department of Computer Systems and Electrical Engineering, Lulea University of Technology on modelling autistic learning and fusion of auditory and visual information in the cortex.He supervised a number of PhD students to graduation. Currently he supervises 7 PhD students.

Abstract:

It is known from psychology and neuroscience that multimodal integration of sensory information enhances the perception of stimuli that are corrupted in one or more modalities. A prominent example of this is that auditory perception of speech is enhanced when speech is bimodal, i.e. when it also has a visual modality.The function of the cortical network processing speech in auditory and visual cortices and in multimodal association areas, is modeled with a Multimodal Self-Organizing Network (MuSON), consisting of several Kohonen Self-Organizing Maps (SOM) with both feedforward and feedback connections.

Simulations with heavily corrupted phonemes and uncorrupted letters as inputs to the MuSON demonstrate a strongly enhanced auditory perception. This is explained by feedback from the bimodal area into the auditory stream, as in cortical processing.

Download the presentation (pdf, 311kb)

Biography:

Ingrid Zukerman is a Professor in Computer Science at Monash University. She received her B.Sc. degree in Industrial Engineering and Management and her M.Sc. degree in Operations Research from the Technion -- Israel Institute of Technology. She received her Ph.D. degree in Computer Science from UCLA in 1986. Since then, she has been working in the Faculty of IT at Monash University. Her areas of interest are discourse planning, plan recognition, question answering, and user modeling.

Abstract:

I will describe a probabilistic approach to argument interpretation that casts the selection of an interpretation as a model evaluation task. In selecting the best model, i.e., that with the highest posterior probability, the formalism balances conflicting factors: model complexity against data fit, and structure complexity against belief reasonableness. Our mechanism receives as input arguments entered through a web interface, and activates an anytime algorithm to produce candidate interpretations.
These interpretations comprise inferences that connect the argument propositions, suppositions that make sense of the beliefs in the argument, and justifications that explain the inferences. Our user evaluations show that the interpretations produced by our system are acceptable, and that there is strong support for the postulated suppositions and justifications.

Download the presentation (pdf, 782kb)

Biography:

I have a Ph.D. from Monash University and shortly after submitting my thesis moved to Silicon Valley where I worked for several companies most notably Silicon Graphics (SGI) as part of their data mining group. In 2002 I moved into Academia as an assistant professor at the State University of New York @ Albany. I publish and serve on the program committees of A.I. and data mining conferences such as AAAI, ECML/PKDD, SIAM DM, ACM KDD etc. We will be giving a tutorial on clustering under constraints at ACM KDD 2006.

Abstract:

Data mining aims to find "novel and actionable" patterns but fail dismally in areas with considerable domain expertise. Consider the problem of clustering (finding distinct sub-populations) within social networks of individuals generated from pandemic simulations. The typical result using standard techniques is that the clusters correspond to groupings along socio-economic and most likely racial lines which is hardly novel to anyone whose lived in the U.S. We and others are exploring the specification of desirable cluster properties (in the form of constraints) and the design of efficient algorithms to find the best clustering given these constraints. We will present complexity results related to this problem, open issues, algorithms and illustrative examples for image facial databases and social networks. This is joint work with Sugato Basu (SRI), S.S. Ravi (SUNY) and Kiri Wagstaff (NASA-JPL).

Biography:

Kim Marriott obtained his PhD in 1988 from The University of Melbourne. He was a research scientist at IBM's T.J. Watson Research Center until he returned to Australia in 1993 to take up a position at Monash University. He currently has an ARC Professorial Fellowship to lead the Adaptive Diagrams Project. His research interests are in constraint-based graphics, computer and human understanding and reasoning with diagrams and other visual languages, constraint programming, and program analysis and optimisation.

Abstract:

Unlike today where the majority of diagrams and textual documents are static, lifeless objects reflecting their origin in print media, the computer of the near future will provide more flexible visual computer interfaces in which diagrams and text adapt to their viewing context, support interactive exploration and provide semantics-based retrieval and adaptation. I will provide an overview of the Adaptive Diagram Research Project whose aim is to provide a generic computational basis for this new type of document.

Download the presentation (pdf, 6.71mb)

Biography:

In a research career spanning more than thirty years, Professor David Green has applied computers to problems as diverse as starfish, bushfires, DNA, and social breakdown. He is editor of the journal Complexity International and author of several books on the new field of complexity.

Abstract:

Networks (nodes joined by edges) provide a simple, but all-embracing model of natural systems. Their universal nature means that important properties, such as critical phase changes and network topology, underlie a vast range of natural phenomena. One of the challenges for researchers in this field is to discover how local interaction and adaptation within networks give rise to global properties. In this presentation, I will explain some of the issues involved and describe the novel insights that provide about such processes as evolution in landscapes, the stability of ecosystems, and the emergence of law and order within social groups.

Download the presentation [Not yet available]

Biography:

Ray Jarvis completed a degree in Electrical Engineering and a Ph.D. in the same department at the University of Western Australia in 1962 and 1968, respectively. After two years at Purdue University he returned to Australia to establish a Computer Science program at the Australian National University (ANU). He was heavily instrumental in the setting up of a Department of Computer Science at ANU in the mid 1970’s and was its first Head of Department. He came to Monash University in 1985 to a Chair in Electrical and Computer Systems Engineering and established the Intelligent Robotic Research Centre (IRRC), of which he is Director, in 1987. The IRRC is now the oldest formally established Centre in the Faculty of Engineering. He is also part of the Vision Systems Engineering Institute, directed by David Suter.

In 2003 he became Director of the Australian Research Council Centre for Perceptive and Intelligent Machines in Complex Environments (PIMCE).

His research interests are Computer Vision and Intelligent Robotics (including aspects of pattern recognition, artificial neural networks, genetic algorithms and automation theory).

Since 1988 he has attracted research funds totalling nearly $10.5m, of which nearly $3.0m are individual grants and the remainder he has led. He became a Fellow of the Institute of Electrical and Electronics Engineers in 1992 and was granted an Australian Research Council Special Investigator’s Award during 1996-1998.

He is currently concentrating his research in a number of Linkage Grant partnerships. One of these is concerned with robotic bush fire fighting and another is about robotic inspection and removal of suspicious abandoned luggage. A third, involving robotic surf rescue, has not yet been decided upon.

Like John Howard, Ray refuses to be drawn on the topic of retirement.

Abstract:

Intelligent Robotics concerns the melding of perception, reasoning and physical actuation in the fulfilment of useful tasks at various levels of automation and some degree of human guidance. The field is rich in sensors, algorithms, paradigms and applications. This talk will construct a brief sketch of the domain and then launch into a ‘show and tell’ mode presentation of case studies drawn mostly from the speaker’s own work.

Download the presentation (ppt, 30kb)

Biography:

Associate Professor Leonid Churilov received a first class BSc (Honours) degree in Operations Research from The University of Melbourne in 1993 and a PhD in Operations Research from The University of Melbourne in 1998. He joined the School of Business Systems, Faculty of Information Technology, Monash University, in 1998 and as of 2006 is an Associate Professor of Accounting Information Systems at the Department of Accounting and Finance in the Monash Faculty of Business and Economics. His area of research expertise is the interface between decision making and process modelling for effective decision support for business and industry. In particular, this involves research projects in optimisation, simulation and systems dynamics, data mining, and knowledge modelling as well as integrated process modelling studies for enterprise planning.

Associate Professor Churilov has undertaken a variety of research and development projects with health care organizations and has won a number of research and consultancy grants. In 2000 the Victorian State Government awarded him a Victorian Fellowship for computer simulation-based research of business processes within acute hospitals emergency departments. For his research on Iso-Resource Grouping in acute health care in Australia, he was awarded a special prize from the Japanese Operational Research Society and this research was nominated as the Australian National Contribution to the program of the 2000 National Conference of Japanese OR Society. In 2005, Dr Churilov's research on process mining in emergency departments was selected as an Australian national contribution to the triennial International Federation for Operations Research Conference.

Abstract:

In this talk we provide a broad overview of how intelligent systems modelling concepts and techniques are used for decision support in the area of health care. The topics will range from the use of data mining for decision support in prostate cancer treatment to multiple criteria decision making for emergency patients triage, and from combining neural networks and simulation for value adding in hospital emergency departments to the use of intelligent techniques for predicting hospital overcrowding. While not discussing the development of new computational techniques, this presentation concentrates on the solutions that are provided to the whole spectrum of real problems faced by patients and/or health care providers thus demonstrating the high relevance of intelligent systems modelling research.

Download the presentation (pdf, 669kb)

Biography:

David Suter holds a BSc in Applied Mathematics and Physics (Flinders), a Grad. Dip. Comp. (RMIT) and a PhD (La Trobe) in Computer Science with a thesis title containing the usual mix of important sounding words relating (both the title and the thesis) to computer vision. He has lectured at La Trobe (Lecturer 1988-1991) and at Monash (Senior Lecturer, Assoc. Prof., Porfessor 1992-present) Universities.
His research activities focuses on topics such as: motion estimation from images (including optic flow), structure from motion, image segmentation, biomedical image analysis, human motion capture and animation, visual tracking, activity detection and classification, face recognition, and the construction of building models from laser scan and image data. He is a member of the ARC Centre for Perceptive and Intelligent Machines in Complex Environments, and the Institute for Vision Systems Engineering (the latter he directs).
Currently he is an associate editor for the International Journal of Computer Vision, and for the journal Machine Vision and Applications (having also just finished a stint as an associate editor for the International Journal of Image and Graphics).

Abstract:

My group is working in the extraction and exploitation of information contained in images. Much of the work is "low-level" primitive information extraction: e.g., what part of the image correspond to the objects of potential interest, how are they moving through the image sequence (tracking), how do you find the same object or person in different views, etc. Some of the work involves higher level information (e.g., face recognition, gait and activity recognition, human pose and motion capture). Moreover, we "stray" into areas such as graphics (human motion and capture), image and film restoration, biomedical image analysis; and we use information from sensors other than cameras (e.g., laser scanners). This talk will give an overview of the work currently carried out by the group.

An overview of the presentation

Biography:

Mark Wallace holds an MA in Mathematics and Philosophy from Oxford University, an MSc Artificial Intelligence from the University of London, and a PhD from Southampton University which was published in 1983 as the book, "Communicating with Databases in Natural Language". Hoping to master a second language, he moved to the European Computer-Industry Research Centre in Munich, where he discovered Constraint Logic Programming (CLP).
Returning to the UK after ten years, he spent another decade at Imperial College London, leading the ECLiPSe CLP team. Besides its use for research and teaching in some 500 organisations worldwide, ECLiPSe is exploited for building commercial software. In particular ECLiPSe is embedded in Cisco's MPLS Tunnel Builder Pro.
His current research interest is in the hybridisation of different techniques and algorithms tailored to large scale industrial combinatorial problems. The aim is to simplify the task of decomposing an optimisation problem into subproblems that can be efficiently solved with different techniques, and to combine those techniques into a tailored algorithm that solves the whole problem.

Abstract:

When solving large scale combinatorial optimisation problems, the best algorithm typically applies different techniques to different parts of the problem. The challenge is to find the right problem decomposition, the right technique to apply to each subproblem, and the right method to combine the techniques so as to generate globally feasible, high-quality solutions. We discuss how the ECLiPSe constraint programming system supports this problem solving process.

Download the presentation (ppt, 485kb)

Biography:

Geoff Webb holds a research chair in the Faculty of Information Technology at Monash University. Prior to Monash he held appointments at Griffith University and then Deakin University where he received a personal chair. His primary research areas are machine learning, data mining, and user modelling. He is widely known for his contribution to the debate about the application of Occam's razor in machine learning and for the development of numerous algorithms and techniques for machine learning, data mining and user modelling. His commercial data mining software, Magnum Opus, is marketed internationally by Rulequest Research. He is editor-in-chief of Data Mining and Knowledge Discovery and a member of the editorial boards of Machine Learning, ACM Transactions on Knowledge Discovery in Data, User Modeling and User-Adapted Interaction, and Knowledge and Information Systems.

Abstract:

Pattern discovery is one of the fundamental tasks in data mining. Pattern discovery typically explores a massive space of potential patterns to identify those that satisfy some user-specified criteria. This process entails a huge risk (in many cases a near certainty) that many patterns will be false discoveries. These are patterns that satisfy the specified criteria with respect to the sample data but do not satisfy those criteria with respect to the population from which those data are drawn. This talk discusses the problem of false discoveries, and presents techniques for avoiding them.

Download the presentation (pdf, 440kb)