70040 research outputs found
Sort by
Multidimensional Index Structure based on Learned Space Filling Curve
No full textDatabase indexing techniques, as a fundamental component of database systems, are an important research area. A well-designed
index can effectively reduce query execution time. However, traditional index models struggle to maintain high performance when
dealing with large-scale datasets. This is because traditional index models cannot adapt past knowledge and experience, and they
cannot easily incorporate additional information such as data distribution and query workload. This issue does not only happen to
one-dimensional indexes but also extends to multidimensional indexes. To address this issue, this paper explores how to leverage
the machine learning (ML) method to construct effective yet efficient multidimensional indexes. We propose a learning-based
approach that enables index structures can adapt different data distributions and query workloads. Additionally, we introduce
dynamic index update strategies to improve update efficiency.
The first study delves into indexing multidimensional point data. We introduce a novel monotonic space-filling curve (SFC)
family and SFC selection method, which allows the index model to select the most suitable space-filling curve based on workload
characteristics and data distribution. Furthermore, we propose a data packing technique that groups SFC-mapped one-dimensional
data, which enables more efficient filtering optimization. Beyond these offline optimizations, we present a multidimensional window
query partitioning method based on the proposed SFC, which allows for efficient query decomposition. This technique effectively
reduces positive candidates while ensuring we can get exact query results during the query process. Finally, we propose a delta
array-based method in our model, which can improve update throughput via delaying index structure updates.
The second study extends the indexing problem to spatial objects, particularly polygon data. Building upon the concept of learned
SFCs, we investigate how to efficiently select optimal SFCs based on the given cost model. Moreover, we improve the heuristic
packing method in the previous work on both construction time and data organization. In addition, we introduce a representative
point selection method to rank polygons, which can overcome the limitation of using only the minimum SFC value as the index
key. This new key design allows the learned index to rapidly determine whether a polygon intersects with a query polygon,
thereby accelerating query execution. To ensure query completeness, we propose a novel query extension method that prevents
missing results when performing spatial polygon queries. Additionally, we explore efficient update mechanisms to ensure the high
throughput of the learned spatial index during data updates.
Unlike the previous two studies, the third research focuses on dynamic index updates. Most existing work on learned indexes
assumes a static data environment, lacking effective strategies to update the index structure when data or query distributions
change. To bridge this gap, we introduce a cost-model-based update mechanism, which evaluates whether the index structure
needs to be updated. Furthermore, we propose a novel update strategy that maintains multiple learned sub-learned indexes as an
alternative update approach. To handle dynamic query workload variations, we discuss a potential method that can proactively
update the learned index model. By dynamically adjusting the learned index structure and update strategy, our method ensures
optimal performance in evolving data environments.
In summary, this thesis systematically investigates learning-based indexing methods for multidimensional point and polygon data
while exploring practical strategies for efficient index updates. Our experimental results demonstrate that the proposed techniques
outperform state-of-the-art methods in various aspects of query performance, including speed, accuracy, and resource efficiency.
These findings provide valuable insights for future research and practical applications of learned multidimensional indexes in modern
database systems
Constant Care: Health Support for Australian Naval Operations 1901-1976
No full textThe Royal Australian Navy (RAN)’s role in advancing the nation’s world-wide interests requires a diverse range of supporting elements ashore and afloat. This thesis fills significant gaps in the existing Australian military medical historiography, by exploring how the RAN’s health services supported its peacetime and warlike operations between the unification of its ex-colonial antecedents in 1901 and the formal instigation of the Australian Defence Force in 1976. The imperative for this thesis pertains to how this historiography has been distorted by its overwhelming focus on the Australian Army. This has neglected how the RAN’s health services supported its peacetime as well as warlike military, diplomatic and constabulary missions, as described by international relations theorist Ken Booth. The ensuing lack of attention has led to a misapprehension within Defence, government and the public that RAN’s health services only need to treat seagoing battle casualties on the same terms as their Army and Royal Australian Air Force (RAAF) counterparts. Yet, Arthur Graham Butler’s seminal First World War official history explained how military health services have three distinct but inextricably linked ‘purposes’: besides providing treatment services, they also entail facilitating operational capability, and returning service personnel to the civilian community. This thesis analyses how and why Butler’s ‘purposes’ (today’s ‘missions’) are relevant between and within military services, in peace as well as in war. To this end it uses three themes, the first pertaining to how the RAN’s health services applied his ‘purposes’ to support its peacetime and warlike operations, and why these needed the same health support. The second theme examines how the imperative to get its ships to sea and keep them there explains why the RAN’s health services ‘operational capability’ purpose differed from its Army and RAAF counterparts, and how this shaped its functions. Finally, it explores why the RAN had a greater imperative for medical interoperability with the Royal Navy surface fleet than the other Australian services until after the Second World War, thence regarding its aviation, diving and submarine capabilities until the British withdrawal from east of Suez in 1971
Electric vehicle charging emissions under different control strategies and temporal resolutions: Case study for Australia
No full textThe increasing adoption of electric vehicles (EVs) presents both challenges and opportunities for reducing greenhouse gas (GHG) emissions. While EVs are essential for decarbonising the transport sector, the emissions from charging vary greatly depending on the generation mix at the time. This study investigates the impact of various EV charging strategies on GHG emissions in different regions in the Australian National Electricity Market (NEM). The study focuses on four key charging strategies–Control Tariff, Timer, Solar Soak, and Convenience Charging. Using real-world data, the analysis evaluates both average and marginal emissions across regions with varying levels of renewable energy integration. Sensitivity analysis showed that coarser temporal resolution in emissions calculations can lead to variances of up to 6.3 %, emphasising the importance of using higher resolution data when available. It was found that the Solar Soak strategy is the most effective in minimising EV charging emissions and can also help with challenges associated with increasing solar exports in the distribution network. The choice between average and marginal emissions intensity factors is also critical in determining outcomes. In Tasmania and South Australia, where renewable energy sources dominate, the use of marginal emission factors resulted in higher EV charging emissions than average emissions due to their reliance on coal and gas as the marginal generators. The sensitivity analysis carried out with emissions data between 2019 and 2023 revealed a negative relationship between renewable energy fraction and emissions intensity and highlighted the importance of aligning EV charging with high renewable generation periods to achieve maximum GHG reductions
Understanding Ethical Implications of AI Enabled Decision Support Systems on the Battlefield
No full textEvaluating the use and safety of prescribed opioid analgesics during pregnancy and pregnancy-related periods using real-world data
No full textThe use of prescribed opioid analgesics, medicines used to treat pain, has risen globally over the past two decades, including among women before, during, and after pregnancy. This increase has raised concerns as the effects of opioid analgesic use during these periods on maternal and infant health remain poorly understood. This thesis leverages real-world data to generate robust evidence on opioid analgesic use and safety during pregnancy and pregnancy-related periods.
Chapter 1 examines the challenges in generating evidence on medicine safety and effectiveness during pregnancy. The chapter explores three key areas: first, how the complexities of medicine safety before, during and after pregnancy impact prescribing decisions; second, the potential of real-world data and perinatal pharmacoepidemiology to address current knowledge gaps; and third, the current evidence regarding opioid analgesic use and safety during pregnancy and related periods.
In Chapter 2, analysis of national dispensing claims (2013-2020) from a 10% random sample of Australians showed that while prescription opioid analgesic use was common among women of reproductive age, most use was short-term, with overall prevalence declining over the study period.
In Chapter 3, analysis of NSW linked health data (2012-2018) revealed higher opioid analgesic dispensing after hospital discharge for caesarean section versus vaginal births and in public versus private hospitals, although only a small proportion of these women exhibited persistent opioid use.
In Chapter 4, a systematic review and meta-analysis of first-trimester opioid analgesic exposure found no substantial increased risk of congenital anomalies, though methodological limitations prevented definitive conclusions. However, the evidence for gastrointestinal anomalies, cleft palate, and atrial septal defect remains inconclusive.
In Chapter 5, a target trial emulation investigating the impact of opioid analgesic use during pregnancy (which our post hoc analyses revealed was mostly short-term) found no meaningful increase in risk for most maternal and neonatal outcomes. However, signals for stillbirth and neonatal death remain inconclusive.
Chapter 6 summarises the thesis findings, evaluates methodological strengths and limitations, and discusses the broader implications for future research, public health and clinical practice. The Chapter also identifies key research priorities to address remaining knowledge gaps about opioid analgesic use and safety in this understudied population
Investigating the diverse actions of mitochondrial uncouplers in cells and db/db mice
No full textMitochondrial uncouplers are small molecules that make mitochondria less efficient so that more nutrients are oxidised to produce a given amount of ATP. Mitochondrial uncouplers are used as chemical tools to study mitochondrial function in vitro and in vivo, and some molecules are in development for the treatment of metabolic diseases. However, one problem in the field is that any molecule that increases proton transport into the mitochondrial matrix independent of ATP synthase can be classified as an uncoupler regardless of off-target activities. Therefore, there are dozens of classes of molecules that are considered mitochondrial uncouplers which exhibit a wide spectrum of phenotypes. Herein we directly compared 15 mitochondrial uncouplers side-by-side in a well-defined cell system to better understand their in vitro dose response profiles and the top molecules with suitable pharmacology and safety profiles were compared in db/db mice.
BAM15 dose-dependently improved body weight and metabolic parameters in db/db mice, with high dose BAM15 treatment yielding statistically significant improvements in body weight, fat pad weight, glucose tolerance, blood glucose, HbA1c, liver weight and triglyceride content. Treatment with high dose ES9 significantly improved glucose tolerance, blood glucose levels, and HbA1c, but increased body weight, liver size and steatosis relative to db/db controls. This same pattern was reflected in mice treated with low dose ES9, though the measured changes in glucose tolerance, blood glucose levels, HbA1c and liver size were not statistically significant
Overall, we found that few mitochondrial uncouplers behaved similarly, with 11 molecules impairing maximal mitochondrial capacity. BAM15 and ES9 had the greatest dose tolerance ranges in vitro. BAM15 had the best overall in vivo effects on body weight, glucose control and liver steatosis in male db/db mice. This thesis highlights the potential of mitochondrial uncouplers as therapeutics for obesity and metabolic disease
An Innovative Approach to Silicon-integrated Surface Modification of High Carbon Steels: Leveraging Waste Silicon and Carbon Sources
No full textThe alarming increase in waste is directly linked to development, as growing populations and expanding economies lead to higher levels of consumption and waste generation. The current "take-make-dispose" model of consumption is unsustainable, depletes natural resources, and contributes to concerning levels of environmental degradation. There is a need for a paradigm shift in waste management, which flips the perspective of the traditional linear model, recognising that one industry’s waste can be another’s raw material. This research is premised on the concept of ‘industrial symbiosis,’ as it presents a detailed investigation of the feasibility of utilising waste materials to enhance the surface properties of carbon steels, addressing the growing need for sustainable materials processing and waste valorisation.
The study focuses on employing waste-derived materials, such as mixed waste plastics (MWPs), spent coffee grounds (SCGs) as alternative reductants, and waste glass as silicon (Si) sources for surface modification. In essence, this project advocates the concept of extracting value from waste, transforming discarded materials into valuable resources for enhancing the properties and extending the lifetime of high-carbon steels. The investigation begins with a comprehensive characterisation of the waste materials using elementary and advanced analytical techniques, including Fourier-transform infrared spectroscopy (FTIR), X-ray diffraction (XRD), X-ray fluorescence (XRF), CHNS elemental analysis, thermogravimetric analysis coupled with differential thermal analysis (TGA-DTG), and off-gas analysis. This detailed characterisation provides valuable insights into the chemical composition, structure, and thermal behaviour of the waste materials, enabling a thorough assessment of their suitability and evaluate their potential as reductants in metallurgical processes. The effectiveness of these waste materials in reducing iron oxide (Fe2O3) to iron (Fe) is systematically assessed using hematite as a model raw material. This comparative evaluation reveals the potential of waste materials as reductants in pyrometallurgical processes, compared to conventional options and establishing their potential for complementing or even replacing traditional carbon sources. The study importantly showcases the synergistic effects of blending different waste materials, realistically imitating a real-time availability of accumulated wastes. It was observed that a combination of MWPs and metallurgical coke (MC) exhibits enhanced reducing capabilities compared to either material alone. This finding emphasises the potential for optimising waste utilisation by strategically combining different waste streams to maximise their synergistic effects.
The research then examines the diffusion behaviour of Si from different sources into high-carbon steel, investigating the impact of the presence or absence of a solid carbon source on the diffusion process. The effects of temperature, carbon source, and their interactions with Si diffusion are analysed, providing a broader understanding of the underlying mechanisms and the factors influencing Si uptake by the steel. Based on the conditions evaluated for Si diffusion into high carbon steel, the research explores the reduction of silica from different sources, including waste glass and SiO₂ powder, using various alternate reductant combinations of MWPs, SCGs, and MC. The influence of different reductant combinations and ratios on the efficiency of Si reduction is studied to determine the optimal conditions for Si formation. The reduction of silica by carbonaceous reductants was achieved through the creation of non-equilibrium reaction conditions facilitated by elevated partial pressures of reducing gases and synergistic effects between waste materials. Compared to their individual usage, a combination of MWP and SCG reductant systems favoured Si formation, as confirmed by phase analyses, emphasising the importance of synergistic effects in waste utilisation. Notably, the study revealed enhanced reduction efficiency to Si from waste glass compared to pure SiO₂ powder when employing the combination of MWPs and SCGs as reductants.
The core of the study lies in investigating the surface modification of low-alloy, high-carbon steels through Si diffusion using waste-derived alternate reductants and Si sources. The effects of different process parameters, such as temperature, reductant ratios, and Si source, are investigated. This systematic investigation aimed to understand the surface modification process with the highest level of Si diffusion into the subsurface of steel. The extent of Si reduction achieved was directly influenced by the quantity, and combination of reductants employed. The liberation of Si was effective in the presence of the combination of reductants- MWP and SCG, MWP and MC. Also, the carbon content within the chosen reductant directly affected the melting behaviour of the steel during the process. The carbon served as a reducing agent, but a significant role was also played in the subsequent incorporation of Si into the steel matrix, likely affecting the final properties of the steel product. Optimisation of the reductants, both in quantity and nature, especially carbon content, was critical for controlling the reduction and Si integration into the steel.
The performance of the Si-integrated steel surfaces is evaluated through a series of functional characterisation tests, including corrosion and wear tests and hardness measurements. The impact of the surface modification on the material’s properties is assessed, and the performance of steels treated with different waste-derived materials is compared. Silicon incorporation onto steel surfaces reasonably improved tribological properties and corrosion resistance. The silicon-modified steel exhibited reduced friction and wear due to the formation of hard and soft phases and demonstrated enhanced oxidative resistance. A substantial increase in corrosion protection efficiency compared to untreated steel was exhibited by the treated steel. This comprehensive evaluation provides a valuable understanding of the functional effectiveness of the waste-driven surface modification approach in enhancing the properties of the steel, demonstrating its potential for improving corrosion resistance, wear resistance, and hardness. By extracting value from waste materials and applying them to enhance the surface properties of high-carbon steels, this research offers a compelling example of how sustainable practices can be integrated into industrial processes. The insights gained from this study pave the way for developing more sustainable and resource-efficient surface modification techniques, offering a promising avenue for the metal industries to reduce their environmental footprint and enhance the performance of their products
Development of MYCN inhibitors to treat high-risk neuroblastoma
No full textNeuroblastoma (NB) is the most common extracranial solid tumour in early childhood, with high-risk NB patients having only 50% chance of survival and experience relapse. MYCN is an oncogene belongs to the MYC family and is a major oncogenic driver of NB. Direct targeting MYCN has been challenging due to its structure with no deep pocket for direct binding of small molecules. Alternatively, indirectly targeting MYCN could be a promising alternative pathway. This study aimed to develop novel MYCN inhibitors that can potentially target MYCN through direct binding to MYCN, reduce MYCN protein expression and stability.
Based on the lead pyrimido[1,2-a]benzimidazole compound (UNSW-SC-36), a novel library of its derivatives was synthesised and tested for their cytotoxicity against MYCN-amplified NB cell lines in the first results chapter. The three most active compounds showed low micromolar IC50 values of 0.05, 0.13 and 0.46 μM against SK-N-BE(2)-C and 0.09, 0.16 and 1.29 μM against Kelly cells, respectively. Two compounds exhibited a small increase in metabolic stability compared to the lead compound (T1/2 = 0.64 min) with a half-life of 1.38 min and 1.23 min, respectively, while one compound displayed a higher microsomal stability of 9-fold than the lead compound with a half-life of 5.63 min. These three compounds decreased MYCN protein level, and their cytotoxicity against NB cells were partially dependent on MYCN expression.
In the second results chapter, the synthesis and cytotoxicity of pyrido[1,2-a]benzimidazoles and pyrimido[1,2-a]indazoles was described. Two compounds possess good potency with the lowest IC50 against SK-N-BE(2)-C cells to be 0.33 and 0.47 μM, and 0.71 and 1.23 μM against Kelly cells. Both compounds had low toxicity against non-malignant human fibroblasts MRC-5 cells (IC50 >20 μM) and decreased MYCN protein levels. They displayed higher metabolic stability than lead compounds (half-life of 3.00 and 0.92 min respectively), and their cytotoxicity effect was partially dependent on MYCN protein expression.
The synthesis of 2-phenylimidazo[1,2-a]pyrimidines and their cytotoxicity were described in the third results chapter with the most active compound showing low IC50 of 0.02 and 0.09 μM against SK-N-BE(2)-C and Kelly cells, respectively, with low toxicity against normal human fibroblast cells MRC-5 (IC50 >20 μM). The most active compounds had similar metabolic stability to that of the lead compound (half-life of 0.62 min). The compound reduced MYCN protein when cells were treated at the respective IC50 values, and its activity was partly dependent on expression of MYCN.
Collectively, this thesis has developed novel small molecule compounds that can be further evaluated in vitro and pre-clinical animal NB models, to provide novel MYCN targeted therapy for the treatment of high-risk NB patients