Filters
Availability
Publication Year
Resource Type
Faculty
School/Institute
Publication Search Results
Sort
Results per page
1 - 10 of 107
-
(2010)ThesisEffective cochlear implant fitting (or programming) is essential for providing good hearing outcomes, yet it is a subjective and error-prone task. The initial objective of this research was to automate the procedure using the auditory nerve electrically evoked compound action potential (the ECAP) and machine intelligence. The Nucleus® cochlear implant measures the ECAP via its Neural Response Telemetry (NRT™) system. AutoNRT™, a commercial intelligent system that measures ECAP thresholds with the Nucleus Freedom™ implant, was firstly developed in this research. AutoNRT uses decision tree expert systems that automatically recognise ECAPs. The algorithm approaches threshold from lower stimulus levels, ensuring recipient safety during postoperative measurements. Clinical studies have demonstrated success on approximately 95% of electrodes, measured with the same efficacy as a human expert. NRT features other than ECAP threshold, such as the ECAP recovery function, could not be measured with similar success rates, precluding further automation and loudness prediction from data mining results. Despite this outcome, a better application of the ECAP threshold profile towards fitting was established. Since C-level profiles (the contour of maximum acceptable stimulus levels across the implant array) were observed to be flatter than T-level profiles (the contour of minimum audibility), a flattening of the ECAP threshold profile was adopted when applied as a fitting profile at higher stimulus levels. Clinical benefits of this profile scaling technique were demonstrated in a 42 subject study. Data mining results also provided an insight into the ECAP recovery function and refractoriness. It is argued that the ECAP recovery function is heavily influenced by the size of the recruited neural population, with evidence gathered from a computational model of the cat auditory nerve and NRT measurements with 21 human subjects. Slower ECAP recovery, at equal loudness, is a consequence of greater neural recruitment leading to lower mean spike probabilities. This view can explain the counterintuitive association between slower ECAP recovery and greater temporal responsiveness to increasing stimulation rate. This thesis presents the first attempt at achieving completely automated cochlear implant fitting via machine intelligence; a future generation implant, capable of high fidelity auditory system measurements, may realise the ultimate objective.
-
(2012)ThesisIn recent years, vision is gaining increasing importance as a method of human-computer interaction. Vision techniques are becoming popular especially within immersive environment systems, which rely on innovation and novelty, and in which the large spatial area benefits from the unintrusive operation of visual sensors. However, despite the vast amount of research on vision based analysis and immersive environments, there is a considerable gap in coupling the two fields. In particular, a vision system that can provide recognition of the activities of multiple persons within the environment in real time would be highly beneficial in providing data not just for real-virtual interaction, but also for sociology and psychology research. This thesis presents novel solutions for two important vision tasks that support the ultimate goal of performing activity recognition of multiple persons within an immersive environment. Although the work within this thesis has been tested in a specific immersive environment, namely the Advanced Visualisation and Interaction Environment, the components and frameworks can be easily carried over to other immersive systems. The first task is the real-time tracking of multiple persons as they navigate within the environment. Numerous low-level algorithms, which leverage the spatial positioning of the cameras, are combined in an innovative manner to provide a high-level, extensible framework that provides robust tracking of up to 10 persons within the immersive environment. The framework uses multiple cameras distributed over multiple computers for efficiency, and supports additional cameras for greater coverage of larger areas. The second task is that of converting the low-level feature values derived from an underlying vision system into activity classes for each person. Such a system can be used in later stages to recognize increasingly complex activities using symbolic logic. An on-line, incremental knowledge acquisition (KA) philosophy is utilised for this task, which allows the introduction of additional features and classes even during system operation. The philosophy lends itself to a robust software framework, which allows a vision programmer to add activity classification rules to the system \textit{ad infinitum}. The KA framework provides automated knowledge verification techniques and leverages the power of human cognition to provide computationally efficient yet accurate classification. The final system is able to discriminate 8 different activities performed by up to 5 persons.
-
(2012)ThesisDescription logics belong to a family of knowledge representation formalisms that are widely used for representing ontologies. However, ontologies are subject to changes and are susceptible to logical errors as they evolve. Ontology reasoners are able to identify these errors, but they provide very limited support for resolving them. In particular, the existing tools do not provide adequate support to prevent logical errors from being introduced into an ontology. In this research, we investigate three different operations that are directly related to the management of logical inconsistencies in ontologies, namely: ontology contraction, integration and debugging. Ontology contraction concerns the removal of information from a set of description logic sentences, where the resulting set of sentences is consistent. Ontology integration is the problem of combining multiple sets of description logic sentences in a consistent manner. Ontology debugging deals with the removal of description logic sentences to restore the consistency of an ontology. In this regard, contraction and integration can be considered as prevention of logical errors, and debugging as cure. We present a construction of contraction for description logics based on the well-known partial meet contraction for belief bases from the area of belief change. We show that this construction produces more refined solutions, and we show that this construction is governed by a refined set of contraction postulates. Moreover, we recast a class of propositional knowledge integration strategies known as adjustments. We show that these strategies cannot be directly used in the description logic setting due to limitations in the expressive power of description logics. We then provide two new adjustment strategies which are appropriate for description logics, and we show that these strategies produce more refined solutions. Furthermore, we study a tableau-based algorithm that identifies the maximally satisfiable subsets (and minimally unsatisfiable subsets) of an ontology. We show that classical blocking do not guarantee completeness in the presence of cyclic definitions, and we provide revised blocking conditions and prove that they preserve both soundness and completeness. Finally, we introduce a diagrammatic approach for debugging ontologies based on Reduced Ordered Binary Decision Diagrams (ROBDDs).
-
(2012)ThesisSpatial and temporal databases play a vital role in many applications in different areas such as Geographic Information Systems (GIS), stock market, wireless sensor network, traffic monitoring and internet applications, etc. Due to their importance, a huge amount of work has focused on efficiently computing various spatial and temporal queries. Among the applications, end-users are more interested in the most important query answers in the potentially enormous answer space. Therefore, different types of information systems use various techniques to rank query answers and return the most important query results to users. Observed in real world scenario, top-k results are more interesting to users and a top-k query is a natural way to be asked to reflect a user s preference in terms of a user-defined scoring function. In this thesis, we provide efficient solutions for the top-k queries under various settings and different criteria for users preferences. Specifically, we tackle three types of top-k queries in a systematic way. Below is a brief description of our contributions. We are the first to study the efficient monitoring of top-k pairs queries over data streams. We present the first approach to answer a broad class of top-k pairs and top-k objects queries over sliding windows. Our framework handles multiple top-k queries and each query is allowed to use a different scoring function, a different value of k and a different size of the sliding window. Furthermore, the framework allows the users to define arbitrarily complex scoring functions and supports out-of-order data streams. We are the first to study the top-k loyalty queries. We propose a measure named loyalty that reflects how persistently an object satisfies the criteria. Formally, the loyalty of an object is the total time (in past T time units) it satisfied the query criteria. We propose an optimal approach to monitor the loyalty queries over sliding windows that continuously report k objects with the highest loyalties. We also experimentally verify the effectiveness of the proposed approach by comparing it with a classic sweep line algorithm. We are the first to study the I/O efficient solution for depth-related problems which can be used for retrieving the top-k objects with linear scoring functions. Half-plane depth of a plane is the number of objects lying in the plane. Location depth of a point p is the minimum half-plane depth of any plane that is bounded by any line passing through p. We propose disk-based algorithms for a few important depth-related problems namely k-depth contour, k-snippet and k-upper envelope. We show that one of our proposed algorithms is I/O optimal for k-snippet and k-upper envelope problems.
-
(2012)ThesisCity streets are full of useful information pertaining to the environment, traffic, commerce and crime, which can benefit the public. Unfortunately, even in today's highly advanced society, most street level information that we desire is still collected manually. Recent advances in mobile phone technology makes it feasible for ordinary users to sense and contribute their ambient information with their smartphones. A new paradigm of PS has emerged, driven by the ubiquitous presence of smartphones. An important form of PS is image sensing, as ``a picture is worth a thousand words''. In this thesis we design and implement a PS system called PetrolWatch that allows volunteers to automatically collect, contribute and share fuel price information. The fuel price is automatically collected from road-side price board images captured by a vehicle mounted smartphone. By leveraging a variety of embedded phone sensors such as GPS receiver and publicly available information in the form of GIS database, our system automatically captures a sequence of fuel price board images when it approaches a service station. These images are then transported to a central server where computer vision algorithms are implemented for board detection and fuel price extraction. Initial road tests show challenges in image capturing and processing. To overcome these challenges, we develop an automatic image capturing system, including a camera control scheme and an image pre-selection algorithm, to capture high quality images from a moving vehicle. We further design sophisticated computer vision algorithms to extract price data from the collected images. Our global colour classification algorithm, based on machine learning techniques, significantly improves the board detection and fuel price extraction rates. We develop a PetrolWatch prototype on a Nokia N95 mobile phone. Extensive driving experiments were conducted in suburban Sydney to validate and optimize our work. Experimental results show that our design achieves a board detection rate of 86%, a price character classification rate of 85%, and a low false positive board detection rate of 10%. The final price extraction rate is 73.1% which combines the board detection rate and price character classification rate.
-
(2012)ThesisWith the discovery that sickle cell anaemia is the result of a structurally altered haemoglobin protein in the 1940s, it was established that genetic variation is the basis of disease. Thus the convention in studying genetic disease became explaining phenotypic variation through causal variations in the genome. Initially, single family-based clinical cases of a disease were investigated, focusing on a single disease-associated region, but this technique was inappropriate for more complex diseases were multiple genes and factors increased the disease risk. Rather, genome-wide and whole genome analyses are performed using thousands of disease cases within a population. These novel techniques in genotyping and genome sequencing are flooding the genetics community with huge amounts of data, but efficient analysis of this information is still lacking. Computational approaches to extract sensible information from genomic data are still in their infancy, with researchers relying on less scalable traditional methods. The aims of this project were to aid in the discovery of candidate disease genes from genome-wide association studies (GWAS) and thus potential molecular mechanisms behind complex diseases by analysing genomic information through data mining. Instead of relying solely on protein pathway analysis as most methods do, we looked at domain homology and regulatory control as tools for predicting likely candidates. Firstly, a method to help identify candidate disease genes and mechanisms in complex diseases that are assayed through GWAS was developed and benchmarked. The methodology was shown to be effective in identifying known disease and disease candidate genes from the GWAS data. Secondly, this method was applied to GWAS on coronary artery disease, hypertension, type II diabetes, bipolar disorder, Crohn's disease, rheumatoid arthritis and type I diabetes. Novel disease gene candidates that were overlooked in the initial analyses were predicted. At most, the number of predictions averaged at around 170 candidates for the average 400 disease-associated loci implicated. The top candidates were validated through literature searches and analyses. Finally, this method was made available to other researchers to use on their candidate disease gene studies from GWAS by upgrading a gene prediction webserver, Gentrepid.
-
(2013)ThesisThe theory of belief revision studies the way in which an agent changes its beliefs as it acquires new information. The changes often involve removal of existing beliefs--the contraction operation--and incorporation of newly acquired belief--the revision operation. The dominant theory of belief revision is the so called AGM framework. This account of belief revision assumes an underlying logic that contains classical propositional logic. Due to this assumption, the framework can not be applied to systems with underlying logic less expressive than propositional logic. This assumption prohibits the usage of the AGM framework to many useful artificial intelligence systems. This thesis aims to remedy this situation by studying belief revision under the Horn fragment of propositional logic which we term as Horn logic. The study extends the applicability of the AGM framework to systems based on Horn logic and it provides theoretical underpinnings for extending the framework's applicability to other non-classical logics. When attacking the problem of belief revision there are two general strategies to follow, that is, to formulate postulates that capture the intuitions of rational belief change and to present explicit constructions of change operations that accord with the postulates. It is widely accepted that the AGM postulates best capture the intuitions behind rational belief change. The AGM constructions of the contraction and the revision operations are shown to be sound and complete with respect to their corresponding postulates. In this thesis, three constructions of the contraction operation is defined for Horn logic. These constructions are studied extensively under propositional logic, the main contribution here is their adaptations to Horn logic. Since Horn logic is a subset of propositional logic, it lacks some logical notions of propositional logic. The main challenge in the adaptation is to compromise the loss of expressivity by using alternative notions and approximation techniques and in doing so preserves the properties of AGM contraction. When we only consider the change mechanisms applied to the beliefs representable by Horn logic, our investigation shows that the adapted contraction operations perform as rationally as AGM contraction in the case of transitively relational partial meet Horn contraction and model based Horn contraction and as a restricted form of AGM contraction in the case of epistemic entrenchment Horn contraction.
-
(2012)ThesisThe Semantic Web is becoming increasingly important for information systems and significant progress has been made in recent years in developing new ideas and tools. As the demand for semantically enabled applications increases, a diverse range of semantic web technologies (SWT) have emerged with the purpose of supporting the development of more intelligent information systems. The evolution and emergence of SWT is mostly fragmented lacking solution blueprints and guidance for organizations in the pathway of adopting SWT to address pragmatic problems. As a result, a majority of the semantically enabled applications in production and development are confined in smaller and trivial use cases. Operations in a controlled and simulated environment are often assumed. In this thesis, we focused on the study of novel semantic-based techniques and tools for addressing the current challenges and needs in domain of Regulatory Compliance Auditing. Regulatory Compliance Auditing involves many pragmatic use cases with effect on broad range of important actors in organizations. These problems could be solved and supported by SWT. For example, if SWT are used to model regulatory policies that relates to the regulations, it would make it simpler and easier for organizations and legislators to add, delete, modify and manage regulations while facilitating audits in an efficient manner. Adopting SWT in Regulatory Compliance Auditing is not without challenges. We have identified and then proposed solutions to address the challenges and issues. Solutions materialized from this thesis include a Semantic-based Reasoning Framework (SemRRF) and an Ontology-based Robust Production System (OntoRPS) that involves SWT such as Ontologies, Semantic-based Rules, Fuzzy Logic and Production Systems. Furthermore, we have described and demonstrated the generic usage and real world application of our proposed semantic-based solution by implementing SemRRF and OntoRPS and presenting a case study for addressing IT security compliance. We aim to use our studies in the Regulatory Compliance Auditing domain as solution guidance for organizations with plans to adopt SWT on more pragmatic, complex and important real-world problems. Another objective is to pioneer research on semantic-based approach to the ubiquitous Regulatory Compliance problems.
-
(2012)ThesisAdvances in computational resources and the communications infrastructure, as well as the rapid rise of the World Wide Web, have increased the wide availability of published papers in electronic form. Digital libraries - indexes collections of articles- have become an essential resource for academic communities. The citation records are also used to measure the impact of specific publications in the research community. Several bibliography databases have been established, which automatically extract bibliography information and cited references from digital repositories. Maintaining, updating and correcting citation data in bibliography databases is an important and ever-increasing task. Bibliography databases contain many errors arising, for example, from data entry mistakes, imperfect citation-gathering software or common author names. Text mining is an emerging technology to deal with such problems. In this thesis, new text mining techniques are proposed to deal with three different data quality problems in real-life bibliography data, which include: 1. Clustering search results from citation enhanced search engines 2. Learning top-k transformation rules from complex co-referent records 3. Comparative citation analysis based on semi-automatic cleansed bibliography data. The first issue has been tackled by proposing a new similarity function that incorporates domain information, and by implementing an outlier-conscious algorithm in the generation of clusters. Experimental results confirm that the proposed clustering method is superior to prior approaches. The second problem has been to develop an efficient and effective method to extract top-k high quality transformation rules for a given set of possibly coreferent record pairs. An effective algorithm is proposed, that performs careful local analysis for each record pair and generates candidate rules, and finally chooses top-k rules based on a scoring function. Extensive experiments performed on several publicly available real-world datasets demonstrate its advantage over the previous state-of-the-art algorithm in both effectiveness and efficiency. The final problem has been broached by developing a semi-automatic tool to perform extensive data cleaning, correcting errors found in the citations returned from Google Scholar, and parsing the citations into structured data formats suitable for citation analysis. The results are then compared with the results from the most widely used subscription-based citation database, Scopus. Extensive experiments on various bibliometric indexes of a collection of research in computer science have, demonstrated the usefulness of Google Scholar in conducting citation analysis, and highlighted its broader international impact on the quality of the publication.
-
(2012)ThesisBenefiting from the recent hardware improvement, multithreaded programs may still introduce concurrency defects which are notoriously difficult to detect, due to the non-deterministic program behavior. One of such defects is known as data race. Programs which have data races often results in inconsistent data and unpredictable behavior. Conventional hybrid approaches of detecting data races suffer from either excessive analysis overhead or precision loss. In this thesis, we introduce three hybrid algorithms for detecting data races in multithreaded programs. By leveraging recent advances on more efficiently tracking the happens-before relation and by developing new lockset algorithms, our detectors can find more potential data races with less false warnings at some slight performance degradation. We first propose Acculock, which is the first hybrid detector that combines lockset and epoch-based happens-before for data race detection. Acculock analyzes a program execution by reasoning about the subset of the happens-before relation, thereby making it less sensitive to thread interleaving than pure happens-before detectors. When this relaxed happens-before relation is violated, Acculock applies a new lockset algorithm to verify the locking discipline, hence making it report less false warnings than the pure lockset detectors. We also propose MPL, which records sets of locksets instead of locksets to detect data races caused by the use of the multiple-protecting-lock idiom, which cannot be detected by Acculock. We finally present an improved version of MPL, MultiLock-HB, which also leverages the epoch-based happens-before technique by combining with a new lockset algorithm to reduce excessive false warnings and to find more races than MPL. All the properties of Acculock have been validated and confirmed by comparing it six other detectors, which are implemented in Jikes RVM using a collection of large benchmarks. Porting Acculock and repeating experiments in a different platform, RoadRunner, yields similar observations and conclusions in terms of their effectiveness in race detection and instrumentation overhead. Replacing the lockset algorithm used in Acculock with MultiLock-HB only suppresses three false warnings in eclipse at the expense of 3X performance slowdown (on average). Therefore, we can selectively apply MultiLock-HB to certain complicated applications where Acculock fails to reason about the multiple-protecting-lock idiom.