Effective Fuzzy Systems for Complex Structured Data Using Fuzzy Signatures. We are developing systematic, heuristic and mathematical techniques to produce effective fuzzy systems for complex structured data. Many or most real world problems have data which has interdependent sub-components depending on the context (eg only female patients need be tested for pregnancy), and often has missing components. Our techniques use fuzzy signatures to extend simple fuzzy systems to deal with data with such ....Effective Fuzzy Systems for Complex Structured Data Using Fuzzy Signatures. We are developing systematic, heuristic and mathematical techniques to produce effective fuzzy systems for complex structured data. Many or most real world problems have data which has interdependent sub-components depending on the context (eg only female patients need be tested for pregnancy), and often has missing components. Our techniques use fuzzy signatures to extend simple fuzzy systems to deal with data with such complex (sub-)structure. This produces effective fuzzy systems with wide applicability to real problems, in telecommunications, and petroleum reservoir data.Read moreRead less
Developing Minimum Message Length and Support Vector Machine methods to predict user behaviour. Predicting and modelling customer behaviour enables considerable savings in the telecommunications industry and elsewhere. The resulting predictive models facilitate identifying novice users, identifying fraud, responding to users' needs, guiding and advising users, and forwarding useful information.
We consider two cutting-edge data mining approaches, Minimum Message Length (developed and led by ....Developing Minimum Message Length and Support Vector Machine methods to predict user behaviour. Predicting and modelling customer behaviour enables considerable savings in the telecommunications industry and elsewhere. The resulting predictive models facilitate identifying novice users, identifying fraud, responding to users' needs, guiding and advising users, and forwarding useful information.
We consider two cutting-edge data mining approaches, Minimum Message Length (developed and led by Monash) and Support Vector Machines, in order to create efficient tailor-made software.
Our software will respond to specific groups of users, and their changes over time, rather than just the average user. Moreover, it will integrate the functionalities of existing individual data mining software.Read moreRead less
Resource-bounded adaptive inference of accurate conditional probability estimates from data. This project will develop machine learning techniques with a valuable new capability: the ability to produce estimates of complex conditional probabilities to varying levels of expected accuracy depending upon the constraints of available computational resources. This will provide significant competitive advantage to developers of many types of online application by allowing them to maximise utilisation ....Resource-bounded adaptive inference of accurate conditional probability estimates from data. This project will develop machine learning techniques with a valuable new capability: the ability to produce estimates of complex conditional probabilities to varying levels of expected accuracy depending upon the constraints of available computational resources. This will provide significant competitive advantage to developers of many types of online application by allowing them to maximise utilisation of available computational resources when making inferences from data, together with the flexibility to trade-off accuracy and computing resources during system design. Australia will also benefit by strengthening its machine learning expertise, which is central to many complex and intelligent systems and the booming data mining industry.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
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.
Transaction Oriented Computational Models for Multi Agent Systems. Agent systems are a very promising technology for constructing complex, large-scale software. Australian researchers have made key
contributions in this area, particularly with reference to one mature and commonly adopted agent architecture known as BDI (Belief, Desire, Intention). To make this technology suitable for use in advanced applications, it has to be provided with robust and predictable behaviour. This project wil ....Transaction Oriented Computational Models for Multi Agent Systems. Agent systems are a very promising technology for constructing complex, large-scale software. Australian researchers have made key
contributions in this area, particularly with reference to one mature and commonly adopted agent architecture known as BDI (Belief, Desire, Intention). To make this technology suitable for use in advanced applications, it has to be provided with robust and predictable behaviour. This project will address that need by designing and implementing a novel agent language for BDI, based on contributions using transactional concepts for agents developed at The University of Melbourne. This will contribute to the development of robust and predictable agent software, that can be used in complex and large scale applications of the future.
Read moreRead less
Sentiment detection from opinion surveys -- the quest for customer and employee satisfaction. The research will yield improved international standing through scientific advances disseminated through high impact refereed publications and open source software. The advances made through the application of sophisticated probabilistic techniques to Language Technology problems will attract post-graduate students, and promote commercial interest. The demonstration prototype will provide proof of conce ....Sentiment detection from opinion surveys -- the quest for customer and employee satisfaction. The research will yield improved international standing through scientific advances disseminated through high impact refereed publications and open source software. The advances made through the application of sophisticated probabilistic techniques to Language Technology problems will attract post-graduate students, and promote commercial interest. The demonstration prototype will provide proof of concept of an application that enables business intelligence to automatically process free-form feedback from customers and employees, with resultant recommendations leading to increased customer and employee satisfaction. The applicability of the outcomes of this research to service industries will further improve Australia's service reputation.Read moreRead less
Query interpretation and response generation in large on-line resources. The unprecedented information explosion associated with the evolution of the Internet makes salient the challenge of providing users with answers to queries posed to Internet resources. The proposed project will apply machine learning and reasoning under uncertainty techniques to leverage the large amount of data found in the Internet in order to perform three tasks: (1) infer users' informational goals from their questions ....Query interpretation and response generation in large on-line resources. The unprecedented information explosion associated with the evolution of the Internet makes salient the challenge of providing users with answers to queries posed to Internet resources. The proposed project will apply machine learning and reasoning under uncertainty techniques to leverage the large amount of data found in the Internet in order to perform three tasks: (1) infer users' informational goals from their questions, (2) modify questions to improve the accuracy of retrieval engines, and (3) compose concise replies from the retrieved documents. The envisioned outcome of this project is a system that will generate replies to questions posed to on-line resources.Read moreRead less
Advanced Bayesian Networks for Epidemiology. We will demonstrate the potential of advanced Artificial Intelligence for medical informatics by extending the capabilities of Bayesian Networks. Bayesian Networks excel when researchers need to combine causal and diagnostic reasoning in areas characterised by uncertainty. But they have one flaw which hinders their use: they do not yet easily mix continuous and discrete variables. We will extend them to handle such mixes, then demonstrate how much the ....Advanced Bayesian Networks for Epidemiology. We will demonstrate the potential of advanced Artificial Intelligence for medical informatics by extending the capabilities of Bayesian Networks. Bayesian Networks excel when researchers need to combine causal and diagnostic reasoning in areas characterised by uncertainty. But they have one flaw which hinders their use: they do not yet easily mix continuous and discrete variables. We will extend them to handle such mixes, then demonstrate how much they can improve on current methods for predicting, among other things, coronary heart disease (CHD).Read moreRead less
Deja-Vu -- A mechanism for constructing dialogue memory for resource-bounded agents. The ability to provide contextual information during interactions with computer systems has great potential to improve the overall experience for users. We propose to develop such an ability in the form of an automatically generated, continuously updated ``dialogue memory'', which may reside at server sites or in the PDAs of individual users. This memory will be generated by means of a novel approach which combi ....Deja-Vu -- A mechanism for constructing dialogue memory for resource-bounded agents. The ability to provide contextual information during interactions with computer systems has great potential to improve the overall experience for users. We propose to develop such an ability in the form of an automatically generated, continuously updated ``dialogue memory'', which may reside at server sites or in the PDAs of individual users. This memory will be generated by means of a novel approach which combines Natural Language techniques to extract dialogue features, model-selection techniques to cluster related dialogues, and cognitive modeling techniques to prune the resultant memories. The implemented computer system will be tested in the domain of trouble-shooting dialogues.Read moreRead less