Open Access Theses
Permanent URI for this collectionhttps://hdl.handle.net/1885/3
To view all theses in this collection, select one of the 'Browse by' options (Issue Date, Author, Title, Subject, Title or Type (of thesis). You can also enter your keyword/s into the text box above and click on Search.
ANU theses are harvested by the National Library of Australia's Trove service and other search engines, making them fully discoverable online.
Find Australian theses.
Full instructions available here
Submit your thesis (Approved ANU research theses only)
Please note: The Abstracts displayed in item metadata are in many cases truncated. For the full Abstracts, see the thesis document files.
Browse
Recent Submissions
Item Open Access Sputter deposition and plasma modification of tungsten alloys for nuclear fusion applications(2025) Mahoney, NicholasThe ITER reactor aims to be first thermonuclear fusion device to demonstrate net fusion power. Critical plasma-facing components in the reactor will be made of tungsten, but helium plasma irradiation is known to embrittle tungsten metal. A future reactor will therefore need an alternative material to extend the lifetime of the components. This thesis investigated the use of tungsten alloy films as a plasma-facing material for a future reactor. Tantalum and chromium were chosen as the alloying elements, and 60 − 100 nm films were manufactured using magnetron sputter deposition at a variety of alloy concentrations. Tungsten alloy films are known to deposit in an undesirable A15 crystal phase. The films were heat treated at 650◦C for 1 hour, 2 hours and 4 hours to induce and investigate the phase transition from A15 to BCC structure. X-ray diffraction patterns confirmed the presence of an A15 structure prior to annealing. The A15 phase was very stable in comparison with past results, requiring over 2 hours at 650◦C to transition to BCC. This was speculated to be linked to oxygen trapped in the films during deposition. The phase transition was slower in alloyed films than a pure tungsten control, with the retarding effect of tantalum stronger than chromium, which suggested the alloying atoms further stabilised the deposited A15 structure. Resistivity of the films saw a reduction with annealing time characteristic of the phase transition. The annealed alloy films were then exposed to helium plasma in the Magnetised Plasma Interaction Experiment at 300◦C, 500◦C and 800◦C. Surface morphology changes were tracked with secondary electron imaging. At 800◦C, surface pits were seen, which were larger in a pure tungsten film than in low concentration alloy films. Advanced surface roughening was observed in the chromium films, while a high concentration tantalum film had no surface damage at all. X-ray diffraction patterns confirmed these trends, and showed an increase in microstrain with the surface roughening in the chromium alloys. These results suggest W-Ta alloy films may have superior irradiation resistance than W-Cr alloy films, but more work is needed to confirm the trends in alloy concentration.Item Open Access Recent Developments in Applying Quantile Regression for Missing Data in Longitudinal Studies(2025) Tao, DehuaIn this thesis, we propose new approaches based on a quantile regression model to handle missing data in longitudinal studies. In Chapters 2 and 3, we develop univariate quantile regression imputation models for longitudinal data that may exhibit population heterogeneity. In Chapter 4, we propose a multivariate quantile regression imputation model capable of handling missing values across multiple responses, as well as imputations for censored observations. Our proposed models are evaluated using three HIV longitudinal datasets collected from the Multicenter AIDS Cohort Study and the Women's Interagency HIV Study (also known as the MACS/WIHS Combined Cohort Study). We believe these models can also be applied to other longitudinal datasets. All our models are built within the Bayesian framework, with estimation and inference implemented using the Markov Chain Monte Carlo (MCMC) procedure. In Chapter 2, considering the presence of population heterogeneity in the longitudinal data, we construct a latent class mixture quantile regression model for imputation. To address missing values under the Missing Not at Random (MNAR) assumption, we propose an additional model to capture the missingness indicator, thereby facilitating the imputation of missing values. We test the proposed approach under various simulation settings, and find the proposed model consistently yields accurate parameter estimations across the simulation scenarios. The model is also evaluated using real data from the MACS study. Based on the real-data analysis results, we conclude that the model fit can be improved with the addition of a latent class structure when a clustering effect exists in the data. In Chapter 3, we extend the imputation model introduced in Chapter 2 by incorporating semiparametric random effects terms, assuming a Dirichlet process prior. This imputation model can be treated as a semiparametric latent class mixture quantile regression model. Additionally, we include the same random effects terms in the model for the missingness indicator, thus building a shared-parameter model structure for MNAR. Simulation results demonstrate that the proposed model can not only provide consistent and accurate results for the estimation of parameters of interest, but can also recover the shape of the random effects terms when the response variable contains missing observations. We also test the proposed model using longitudinal data from the MACS HIV study. Compared to the model presented in Chapter 2, the Chapter 3 model returns a better fit. In Chapter 4, we propose a multivariate imputation approach within the quantile regression framework to address missing values in multiple responses within longitudinal studies. This model is motivated by real data from the WIHS HIV study. The response variable, HIV viral load, contains missing values and is also subject to upper detection limits, which can be treated as left-censored observations. Other variables of interest, such as CD4 and CD8 cell counts, also contain missing observations. Additionally, we incorporate an autocorrelated error structure in the imputation model, as current values can be influenced by previous observations. We test the model under various simulation settings and show that our model can provide accurate and consistent parameter estimation results by imputing the missing observations and recovering the censored observations. In the real data setting, we assess our model's performance compared to models without autocorrelated error structures and those using multivariate linear mixed effects models as imputation models, and we find that our model yields a better fit. In Chapter 5, we summarize the thesis and outline directions for future research.Item Open Access Three essays in Corporate FinanceCui, HeddaThis thesis consists of three chapters. The first chapter is titled "The Glass Cliff: Are Promotions to CEOs Precarious for Women?" This study investigates the "glass cliff" phenomenon, wherein female CEOs are disproportionately appointed to precarious leadership positions. Our analysis demonstrates that female CEOs are approximately 19\% more likely than their male counterparts to be appointed under such challenging conditions. We empirically test two competing hypotheses for the glass cliff: discrimination and preference. We find that the sensitivity of female CEO appointments to firm performance varies with product market competitiveness; it diminishes when industry competition intensifies. Contrary to the preference hypothesis, which suggests that firms may appoint female CEOs for performance recovery, our finding does not support superior post-appointment performance for firms led by women compared to those led by men. Our further analysis shows that while the presence of female directors on the board is associated with an increased likelihood of female CEO appointments, it does not mitigate the glass cliff effect, suggesting that boardroom gender diversity alone is insufficient to address deeper issues related to gender bias in leadership appointments. Overall, our findings imply that the glass cliff is primarily a consequence of gender discrimination rather than a strategic preference for female leadership during precarious situations. The second chapter is titled "Corporate Governance and Performance of Australia Family Firms". While the performance of family firms has been extensively studied in international markets, there is limited understanding of their corporate governance structures and impact on firm performance within the Australian context. In this study, we analyse a large sample of Australian firms to assess the performance of family firms from a corporate governance perspective. We first compare key governance mechanisms between family and non-family firms, including board independence, board size, managerial power concentration, and overall governance quality. Our findings reveal a significant substitutive relationship between family control and conventional governance mechanisms, indicating that family firms often rely on internal governance practices rather than external oversight. When examining performance, we find that family firms exhibit higher profitability; however, this advantage diminishes for market-based valuation once governance heterogeneity is considered. Our results remain robust across various tests, including propensity score matching (PSM) and instrumental variable (IV) analyses. The third chapter is titled "Bank CEOs' Financial Distress Experience and Risk Inertia in Risk-Taking". This study investigates the relation between banks' risk-taking behaviour and bank CEOs' experience in financial distress under regular market conditions. We document strong evidence showing the relationship to be positive: banks of high-risk levels are associated with CEOs with distress experience. This finding is in contrast to previous studies that report a negative influence of CEOs' experiences in crises or early career disasters. Our results remain robust through propensity score matching (PSM) and are further validated with placebo tests. To understand the mechanisms behind the positive relationship, we compare risk levels before and after the appointment of distress-experienced CEOs. Our results show that bank risk-taking does not significantly change following the appointment, suggesting that these CEOs do not actively escalate risk. Instead, the evidence suggests a role of matching in top-management appointments, whereby risk-prone banks strategically choose CEOs with distress experience to align with their existing risk culture.Item Open Access Genomic epidemiology of oat crown rust disease in Australia and abroad uncovers diverse mechanisms to generate diversity(2025) Henningsen, EvaRust fungi are a diverse group of plant pathogens consisting of over 8,000 species in the Basidiomycete order Pucciniales. They parasitize numerous plants of agricultural and ecological importance, such as cereals, legumes, and trees. Like plants, rust fungi undergo alternation of generations during their life cycle. This means that their cells either have one haploid nucleus or two haploid nuclei (dikaryon) at different stages of reproduction. Rust fungi are typically dikaryotic during asexual reproduction, which can occur indefinitely through infection by urediniospores. Host adaptation by mutation during asexual propagation may occur rapidly in rust fungi, but diversification is further accelerated when the sexual cycle is active. The dikaryon enables a third mechanism for generating diversity in the absence of the sexual cycle, in which two individuals exchange entire nuclei without recombination (nuclear exchange). However, it is not understood how much these three mechanisms may contribute to host adaptation at a population scale in most rust species. A pathosystem which is well-suited to explore this is Puccinia coronata f. sp. avenae (Pca), which causes crown rust of oats. Pca is a heteroecious rust which infects oat (Avena spp.) as its telial host and buckthorns (Rhamnus spp.) as its sexual host. Compared to other cereal rusts, Pca is unique in being able to infect globally distributed wild grass species which serve as a reservoir for inoculum. In addition, different modes of reproduction vary in their relevance to Pca populations from different regions, as Rhamnus species are not distributed globally. Australia is a major oat exporter and in epidemic years Pca represents a constraint to production, but the population in Australia has not been characterized with contemporary molecular methods. To characterize the Australian population, Pca samples from across the country were phenotyped on two oat differential sets and their genomes were sequenced with short reads. Previous reports of high phenotypic diversity were confirmed but a high number of lineages were detected, which was unexpected considering the previous assumption of clonality. Chromosome level genome references for diverse Pca isolates from the USA, Australia, South Africa, Egypt, Israel, Switzerland, and the United Kingdom were generated to build resources for characterizing a broader set of Pca isolates. At the haplotype level, previous reports of genetic recombination in the USA population were confirmed and either sexual or cryptic recombination between Australian isolates was detected. The recent invasive ornamental shrub R. alaternus could be acting as a sexual host for Pca in Australia, but this requires further validation. Nuclear exchange events were identified between Pca isolates that are not only associated with significant changes in fitness but also imply intercontinental migration of haplotypes, which provides further impetus for molecular monitoring of rust pathogen populations on a global scale. The haplotype atlas generated through this work represents a significant resource for exploring how virulence evolves in Pca.Item Open Access Bone decomposition in aquatic environments(2025) Macfarlane, NicoleNumerous studies have been undertaken on decomposing human remains in different terrestrial contexts. Determining postmortem interval and bone decomposition rate in cases where remains have been located in aquatic contexts, however, has been problematic due to numerous variables in the aquatic environment. These include but are not limited to water temperature, chemistry, bacteria, scavengers and current or tidal activity. Hence, research in this field has occurred in a somewhat limited capacity, with studies mainly focusing on the effects of water temperature and microorganisms on the decomposition of remains. This study used a controlled animal model experiment over 12 months between January 2019 and December 2019 in the Gold Coast region of Australia. Forty-eight de-fleshed pig femora were placed in tanks replicating differing types of aquatic environments, including riverine, oceanic, estuarine and dam, with a further four femora situated terrestrially to act as controls. Twelve femora were placed in each water type, with three removed at intervals of three, six, nine and twelve months, and four femora on a gravel surface (terrestrial control) surrounded by a cage, with one removed at the previously mentioned intervals. All femora were recorded before and during the fieldwork to examine the effects of taphonomic variables such as individual environment, water chemistry, marine scavenger activity and weathering. Following the removal and collection of the femora at the three-, six---, nine--- and twelve- month intervals, additional analysis was undertaken on the remains, including macro and microscopic quantification of bone surface modifications. This analysis, including measuring any increase in porosity to the surface of the bones, assessing changes in the weight of the bone from commencement to completion of the study, measuring the staining to the interior surface of the bone once cut into a thick section, and histologically inspecting and measuring any inclusions and infiltrations to the thin sections of bone, was used to determine the potential use of each in estimating bone decomposition rate in aquatic environments. The results demonstrated that water chemistry in differing locations plays a large part in the way in which remains decompose. For example, bones subject to the replicated ocean and estuarine environments were slower to decompose due to the scale-forming nature of the water, whereas bones subjected to the more corrosive environments of the riverine and weir environments were more rapidly degraded. It was found that it is indeed possible to measure and quantify bone diagenesis in differing aquatic environments and formulate methods to determine how long remains potentially may have been exposed to water, a bone decomposition, rate using both macroscopic and microscopic techniques. The implications are that in forensic cases where an isolated bone is located in an aquatic environment, researchers may be able to better determine how long this bone had been exposed to water in situ using the techniques developed with this research.Item Open Access Actor and Factor Dynamics in Urban Waste Management Experiments: Insights from China(2025) Wang, ShengnanEffective municipal solid waste (MSW) management is critical due to its significant environmental and public health impacts. Cities play a key role in addressing MSW challenges, while generating the most waste and facing the greatest impacts, with their resources and ability to coordinate stakeholders, they are well-positioned to implement swift, transformative waste management solutions. However, MSW management in cities is a "wicked problem," marked by complex stakeholder interactions, multifaceted challenges, and uncertainties.The concept of sustainability transitions offers a valuable theoretical framework to examine the intricate processes by which MSW management can evolve. It highlights the interactions between actors and systemic factors across multiple levels of governance and practice. Given the uncertainties surrounding both the nature of the problem and potential solutions, small-scale experimentation has emerged as a promising approach to foster sustainable practices. While sustainability experiments are increasingly driven by diverse actors, there is limited understanding of how their roles, interactions, and strategies evolve across developmental stages of sustainability experiments formed in different settings. Moreover, the critical factors and their joint impact that influence outcomes remain poorly explored. Using China,the world's largest producer of MSW and characterized by a top-down governance system, as a case study, this research examines the development of waste management experiments within a centralized policy framework. We explore the primary question: How do actors' roles and strategies evolve in sustainability experiments, and what critical factors and synergies determine their outcomes within China's socio-political and institutional framework?Through four interrelated chapters, this research systematically examines the dynamics of actor roles, interactions, and contextual factors in forming urban waste management experiments across varied settings. The findings demonstrate that non-governmental organizations (NGOs), municipal governments, businesses, and inter-organizational alliances play critical roles in mobilizing resources, strengthening networks, aligning stakeholder interests, and building long-term linkages to address waste challenges collaboratively, albeit their roles and influence evolve over time. This research also identifies three successful causal pathways that underpin effective waste experiments:1. dedicated participants supported by strong community engagement, 2. committed actors backed by local government involvement, and 3. robust grassroots leadership enabled by responsible and responsive local governance. These pathways underscore the critical interplay of social, organizational, and behavioral factors in addressing the complexities of MSW management.By focusing on China's urban waste management, this PhD research makes several key contributions to the field of sustainability transitions. First, it advances understanding of how diverse actors and their interactions evolve across different stages of sustainability experiments, providing nuanced insights into actor dynamics and strategies. Second, it identifies and contextualizes the critical factors and their synergies that drive successful outcomes in sustainability experiments. This research also expands the geographical scope of sustainability transitions research, addressing a notable gap in understanding experimentation within China's unique institutional and socio-political environment. Finally, this work offers practical guidance for policymakers, practitioners, and researchers aiming to design and implement effective waste management initiatives in complex urban contexts.Item Open Access Positron Scattering From Small Hydrocarbons(2025) Cheong, ZoePositrons are anti-matter electrons and have applications in testing of fundamental physics and non-invasive imaging. Positrons can also bind with electrons to form positronium atoms, Ps, and can also bind to atoms and molecules. A phenomenon related to positron binding is scattering resonances, where an incident charged particle is temporarily bound to a target molecule. Contrary to electron scattering, there is limited experimental evidence for positron resonances, only the Surko group in UC San Diego have managed to observe Feshbach resonances via the annihilation cross section for positron scattering from hydrocarbons. This PhD aimed to search for more evidence of the Feshbach resonances for molecules and generate new cross sections for these molecules as well. The resonances observed by the Surko group are vibrational Feshbach resonances (VFR). There is a theoretical framework developed that uses the positron binding energy as a free parameter and assumes that the annihilation escape channel for the VFR decay is small compared to other channels, and it predicts the VFR structure in the annihilation cross section well. However, a corollary of this framework is that the resonance structure can couple to vibrational states lower in energy and allow the positron to escape the resonance structure, known as intramolecular vibrational redistribution (IVR), which is information that the annihilation cross section is unable to provide. The hydrocarbons chosen for the experiment in this PhD were ethane, ethylene and acetylene, which are all two-carbon molecules with an increasing number of bonds between the two carbon atoms. The measurements in this PhD set out to conduct were primarily the vibrational excitation cross sections to search for evidence of inelastic escape of positron from the VFR structure as well as the effects of IVR. Despite strong resonance structure in the annihilation cross section associated with the CH stretch mode of all three target molecules, the CH stretch excitation cross section show no enhancement in the cross section for all three target molecules. The CH3 deformation mode cross section for ethane shows no resonance enhancement either, while the CH2 deformation mode for ethylene shows evidence of multimode VFR structure. The CC stretch mode for both ethylene and acetylene are dipole forbidden, and that of ethylene has a very small cross section and it is unclear if it is real, while that of acetylene shows a clear resonance structure that can only be a VFR coupling into this vibrational mode via IVR. The total cross section for positron scattering from ethane, ethylene and acetylene for this energy range was also measured and showed no resonance structure in the cross section. Consequently, it can be interpreted that for these three molecules, the VFR does not decay via elastic escape. This, combined with the limited evidence for inelastic escape is contrary to the assumptions used by the theoretical calculations. While there is evidence for IVR, it remains that the theory needs to be improved upon. Next, the differential cross section (DCS) of all three target molecules were measured, and are all strongly forward peaked. Such behaviour is typically attributed to molecules with a dipole moment, which ethane, ethylene and acetylene lack, and the strong forward scattering must then be due to the polarizability of the target molecules, which is not accounted for well by the theory. The current grand total cross sections measured were all consistently around 10 angstrom squared larger than the previous measurements, which is attributed to the previous experiments missing a larger portion of the cross section due to the strongly forward peaked DCS. The total elastic, total electronic excitation, positronium and ionisation cross sections are presented as well, where the theory again misses the forward scattering and is smaller than the experimental total elastic cross sections.Item Open Access Machine Learning Assisted Signal Enhancement(2025) Yan, Longfei FelixIn an era of abundant signals, the ability to obtain desired signals while rejecting undesired ones has become increasingly crucial. Often, the desired signals are mixed with interference or contaminated by noise. Signal enhancement techniques play a vital role in performing tasks such as signal separation, extraction, and suppression. This thesis addresses critical challenges in signal enhancement tasks by harnessing the power of machine learning techniques. Firstly, we propose a novel independence criterion called the Finite Basis Independence Criterion (FBIC). This criterion estimates the Hirschfeld-Gebelein-Rényi maximum correlation coefficient between tested variables and is based on mapping functions from a subspace of finite basis. FBIC detects dependence between variables in linear time and outperforms more computationally expensive kernel-based counterparts. Extensive testing in Independent Component Analysis benchmarks demonstrates its potential for various signal separation applications. Secondly, we conduct a comprehensive robustness analysis of a popular signal enhancement approach: fixed beamforming based on first-order linear Differential Microphone Arrays (DMAs). We demonstrate that both bounded and unbounded phase errors of microphones can affect the mainlobe orientation of the beamformer. Analytically derived white noise gain thresholds indicate when mainlobe misorientation occurs. Through rigorous mathematical derivations, we prove that a higher number of microphones and increased spacing between microphones contribute to the robustness of the beamformer. This work provides practical guidelines for designing robust first-order linear DMAs. Thirdly, we propose a neural network model to optimize both the geometry and spatial filter of linear DMAs. The model consists of two feed forward neural networks and is trained end-to-end. The signals enhanced by this model exhibit superior quality compared to those obtained from conventional DMA approaches. Furthermore, the model offers flexibility in controlling the tradeoff between different performance metrics, allowing for customized optimization. Lastly, we extend the neural network model to a general framework that allows optimization of microphone arrays of any geometry, along with their spatial filters. This model employs ResNets and augmented Lagrangian techniques to achieve state-of-the-art frequency-invariant fixed beamforming performance. We showcase our performance in linear, circular, and concentric circular microphone arrays. Moreover, our findings challenge the conventional belief that concentric circular arrays require multiple rings, as we demonstrate that good performance can be achieved with only one ring. Overall, this thesis contributes novel techniques and insights to the field of signal enhancement, leveraging machine learning approaches to address key challenges. The proposed criteria, guidelines and models have the potential to advance various signal separation applications and enhance the overall quality of processed signals.Item Open Access Timor-Leste's Colonial Heritage from the Early Modern Era to the Second World War(2025) Fahey, AndreaThe study of Timor-Leste's colonial heritage requires analysis over four centuries of history. The country has been historically underestimated despite its limited resources, resilience and agency. Traditional academic views influenced by a military-oriented framework fail to explain how Timor-Leste survived and maintained its culture under imperialism and colonialism. Timor-Leste has proven to be a pivot in Asia, an actor to be watched, whose circumstances have had consequences elsewhere. This study adopts a small states perspective and a multilingual methodology to understand the strategic value of Portuguese Timor and explore its unique challenges of state formation from the early modern era up to the Second World War.Item Open Access Theorising the Dingo Howl: Jazz in Australian Film in the 1990s and Early 2000s(2025) Tsuei, JeremyIn discussing the films Dingo (1991), The Boys (1998), Come in Spinner (1990), and The Tracker (2002), I examine the meeting of jazz music as a “sonic indicator of authenticity” with the critical notion of settler-colonial Australia as an unreconciled, uncanny, and inauthentic space. These films tell uniquely Australian stories and diegetically or non-diegetically feature improvised music practices that fall under the category of jazz as a broad “coverall” term. In these films, the music unsettles us and portray an uneasy relationship between their Australian subjects and the Australian landscape. They are also unique films in how their music and musicians were respected and prioritised, sometimes with significant implications for how scenes were filmed and composed. I argue that these films are significant in reflecting the cultural tensions at the time they emerged, many which can still be felt today. These films connect with contemporary developments in jazz practice in Australia and offer a template for future practice-based research and cultural collaborations.Item Open Access Linear and Non-Linear Polarising Meta-Optics Enabled by Photonic Inverse-Design(2025) Li, NeutonOptical devices are ubiquitous in many of the most advanced technologies that we have developed, from polarisation cameras, to space-faring telescopes, to nanostructured displays in augmented reality headsets. Metasurfaces have the potential to disrupt and revolutionise all of these systems, by offering a compact solution with unmatched novel functionality. Whatever the promise of metasurfaces, it is still a challenging effort to design efficient and robust devices. In this thesis, we formulate and develop a novel and general inverse design optimisation scheme that enables the freeform metasurface topologies. This massively expands the utility of our optimisation algorithm, which we then employ to several different applications covered within this thesis. Our primary focus are applications which involve manipulating multiple properties of light simultaneously. These aim to address the most compelling research goals, from high efficiency polarised diffraction, to polarised phase derivative imaging. We first explore the intersection of interactions between polarisation, diffraction, dispersion, and incident angle. These understandings lead us to optimise for different scenarios that would have the highest benefit using metasurfaces. We report the first single metasurface that has beam splitting into the same polarisation states, demonstrating high conversion efficiency of unpolarised light. We also report a novel idea of continuously tunable polarisation dichrosim by varying the tilt of the metasurface, allowing the metasurface to act as fundamentally different polarisers simultaneously. In both examples, we fabricate and experimentally validate our simulated performance with optical characterisation. Next, we explore more complex systems involving multi-objective optimisations. We propose dispersion engineered beam splitting metamaterials, which is achieved by optimisation over the entire metamaterial volume. Our design pushes the boundary on what is possible for broadband devices. Another application is analogue optical processing using a single piece of metasurface. We achieve the largest asymmetrical NA linear OTF for phase derivative imaging, as well as single-shot quantitative imaging using polarised sensing. We believe that our advancements in these metasurface technologies will lead to future novel optical devices with practical utility. Finally we extend our inverse design optimisation algorithm for non-linear metasurfaces, especially the case of sum frequency generation (SFG) for this thesis. Non-linear metasurfaces are an attractive option for creating entangled photon sources, as well as compact up-conversion devices. We employ our optimisation algorithm to design freeform metasurfaces with SFG that have purposeful diffraction shaping and polarisation sensitivity, all the while maintaining high generation efficiency. We anticipate that such inverse algorithms will be crucial in designing metasurfaces for complex applications, such as non-linear quantum photon sources.Item Open Access Learning Multi-modal Fusion for RGB-D Salient Object Detection(2025) Song, PeipeiSalient objects refer to the conspicuous objects within an image that stand out prominently from their surroundings. Saliency detection can be applied in multiple fields, such as object recognition, image compression, and video surveillance. However, it is challenging to detect salient objects from RGB images in cases, such as low illumination or bad weather. Depth maps (D) are commonly utilized as supplementary inputs to RGB images for Salient Object Detection (SOD), namely RGB-D SOD. Although many RGB-D SOD methods have explored various multi-modal fusion strategies for boosting saliency performance, there still exist unsolved problems. In this thesis, we focus on three problems: the first problem is low-quality depth. Due to the diverse acquisition sensors, such as infrared detectors and stereo cameras, the quality of acquired depth maps is inconsistent. The low-quality depth introduces noise that may seriously reduce the detection accuracy. The second problem is the limited receptive field of conventional Convolutional Neural Networks (CNNs). CNN-based multi-modal fusion strategies fail to extensively model the correlation between the two modalities (appearance information from the RGB image and geometric information from the depth data). The third problem is ineffective feature fusion for multi-modal data. Selecting helpful information from each modality is essential to ensure effective data fusion. In this thesis, we introduce three approaches to tackle the problems above. Firstly, we propose a triple attention framework based on a 3D CNN. Triple attention leverages multi-modal features at three levels: modalities, channels, and spatial positions. The modality attention learns the quality factors based on the overall modal features. The channel attention highlights features in the dimension of channels, and the patch-level spatial attention establishes long-range dependencies across the entire image. Using attention mechanisms enlarges the limited receptive field in CNNs and obtains a comprehensive understanding of the context. Existing SOD studies neglect quality assessment of depth maps. To address this gap and enable the evaluation of SOD methods across diverse depth quality levels, we propose a novel quality assessment criterion for depth maps to re-categorize existing RGB-D datasets into three levels: high-quality, mid-quality, and low-quality. Secondly, due to the great success of transformer networks in visual tasks, we propose a transformer-based multi-modal fusion module for saliency detection. Traditional CNNs process images in a hierarchical manner, while the Transformer treats an image as a sequence of tokens and captures long-range dependencies and relations across the entire image. The self-attention mechanism can enhance the unique features of each modality, and the cross-attention can effectively model the correlation between the RGB image and depth map. By combining self-attention with cross-attention in our proposed parallel structure, our model can effectively explore the contribution of each mode for RGB-D saliency detection. Thirdly, we focus on the variant of stereo image pair-based saliency detection, where the depth is implicitly encoded in the stereo image pair for effective RGB-D saliency detection. Compared with RGB-D models, it does not suffer from the modality gap or the low-quality depth issue. Instead of explicitly using the depth map, we propose to model the correlation between the stereo image pair to capture geometric information implicitly using the network. Channel-weighted attention and depth-aware feature grouping are proposed for stereo feature fusion. In summary, we propose three SOD approaches. Comparison results with state-of-the-art methods demonstrate the effectiveness of our proposed approaches. Extensive ablation studies are conducted to validate the contributions of each proposed approach in our work.Item Open Access Characterising two P. falciparum transporters essential for life cycle progression(2025) Huppertz, FrederikDuring life cycle progression, Plasmodium parasites rely on a complex repertoire of transporter proteins to supply them with nutrients obtained from their host. Of the 144 transporter proteins annotated in Plasmodium falciparum, the majority has already been deemed likely essential for asexual blood growth. In this thesis, I hypothesized that a part of the unassigned parasitic transportome will be specifically essential to the sexual gametocyte stages and thus could contain transmission blocking targets for future application. The transportome of P. falciparum was screened in silico for potential candidates and the most promising genes were then targeted via a Crispr-Cas9-based disruption approach. Through this, I was able to confirm a defect in gametogenesis caused by disruption of PfGEP1 previously observed in P. yoelii. Both male and female gametocytes lacking GEP1 were unable to produce gametes. Overall, my data indicates that GEP1 plays a central role in the activation process, making it a promising transmission blocking target. Disruption of a second candidate gene called MCP2 did not affect P. falciparum asexual and sexual blood stages. Deletion of its P. berghei ortholog instead resulted in impaired sporozoite formation in the mosquito vector. While parasites lacking PbMCP2 were able to infect mosquito midguts and form oocysts, only few of them appeared to produce midgut sporozoites. Consequently, very few sporozoites colonized mosquito salivary glands. While lack of MCP2 does not seem to affect vertebrate-to-mosquito transmission as anticipated, it appears necessary for life cycle progression. Together, the findings presented here underline the importance of transporter proteins for Plasmodium parasites throughout their life cycle and contribute to our understanding of their roles in the different life cycle stages of malarial parasites.Item Open Access Implicit Mixtures of Experts for Rigorous Interpretable Machine Learning(2025) Elazar, NathanThis work aims to provide a rigorous framework for practical Interpretable Machine Learning (IML). While IML research has exploded in popularity in recent years, the vast majority of works in the field propose no definition of what interpretability actually means, and even fewer propose practical objective metrics which can quantify the interpretability of models. In this work I define interpretability to be inversely proportional to complexity. While it is difficult to compare the complexities of two arbitrary functions, there are certain model classes which do permit easy measurement of complexity, most notably: the complexity of a decision tree can be measured by the number of nodes it contains. Therefore, as long we restrict ourselves to only considering models from a complexity-measurable class, interpretability is well defined and objectively quantifiable. Unfortunately, for many tasks decision trees perform sub-optimally even when they are allowed to be arbitrarily large. For these tasks, we would like to be able to use more powerful models, such as neural networks. The major technical contribution of this work is the development of the Implicit Mixtures of Experts via Neural Networks (IMoENN) methodology. IMoENN uses neural networks to implicitly generate Mixture of Experts (MoE) models. While explicitly training a MoE directly is only feasible for small mixtures, IMoENN can produce implicit mixtures with arbitrarily many experts. In addition, IMoENN's prediction performance generalizes as well as the neural network architectures that it employs, whereas explicit MoE are prone to overfit to the training dataset. Even when using extremely simple experts, such as linear functions, IMoENN can match deep neural network classification accuracy on image benchmark datasets MNIST10, Fashion-MNIST10 and CIFAR10, provided that there are sufficiently many experts in the mixture. If the experts that make up a mixture are simple models, such as linear functions or decision trees of fixed size, then the complexity of a MoE is readily given by the number of experts in the mixture. In addition, when using IMoENN, the accuracy of these mixtures increases up to that of a black-box classifier as the number of experts is increased. With these facts in mind, I argue that MoE are the ideal model class on which to base complexity-grounded IML. In this work I will demonstrate how IMoENN can be used to create mixtures of linear experts which match black-box accuracy on MNIST10, Fashion-MNIST10, and CIFAR10. These mixtures provide local interpretability by showing the expert responsible for classifying a particular data point. The quality of that local interpretation can be measured by the number of experts in the mixture. IMoENN can also be used for global interpretability, so long as the mixture is small enough that every expert can be inspected. When using such small mixtures IMoENN is unable to match black-box accuracy, so I propose two variants of IMoENN which exploit properties of natural images to improve performance. These variants can match black-box accuracy on MNIST10, but are still worse on Fashion-MNIST10 and CIFAR10. Finally, I show how IMoENN can be used to both learn semantically meaningful features from data and assign feature importances to those features at the same time. This method can provide a degree of global interpretability in the form of global feature importances, and can match black-box accuracy on MNIST10 and Fashion-MNIST10. While I have focused on image classification benchmarks in this work, IMoENN is a very general methodology and can potentially be applied to any task. Therefore, IMoENN is a promising approach to making complexity-grounded IML practically viable on complex datasets, thereby providing a much needed objective metric of interpretability.Item Open Access Essays on Private Health Insurance in Australia: Household Decisions, Participation Projections, and Policy Explorations(2025) Li, Qingyue (Marco)This thesis presents several research topics around Private Health Insurance (PHI) in Australia. Australia's healthcare system features a unique blend of public and private health services, with PHI playing a crucial role in supplementing the publicly funded Medicare system. The participation in PHI is shaped by a range of factors, including government policies, demographic shifts, and economic conditions. Chapter 1 provides a comprehensive analysis of these factors, investigating what drives Australian households' decisions regarding PHI. Subsequently, Chapter 2 introduces a dynamic microsimulation model designed to project future PHI coverage in Australia, incorporating the adverse selection effect. Finally, Chapter 3 employs this model to evaluate the impacts of various assumptions, scenarios, and potential policies on PHI participation and sustainability. Chapter 1 presents an empirical analysis of the determinants influencing households' decisions to acquire or cancel PHI. The analysis highlights the effectiveness of government policies such as the Medicare Levy Surcharge (MLS) and the Lifetime Health Cover (LHC) in encouraging PHI uptake. It also reveals that younger individuals are more likely to cancel their cover or less likely to purchase a policy, while those with higher financial literacy and less financial risk aversion are less likely to drop their PHI coverage. Chapter 2 details the development of the dynamic microsimulation model, which forecasts future trends in PHI participation. The model accounts for demographic changes, family formation patterns, transitions in education levels, and income variations. Crucially, it includes the adverse selection effect, where premium increases from the previous year influence households' PHI decisions in the current year. The model forecasts a significant decline in PHI participation due to adverse selection, with a diminishing proportion of young people among policyholders. Chapter 3 further explores the use of the microsimulation model to examine the poten- tial outcomes of adjustments to key variables, such as mortality rates and wage growth rates. It also explores variations in the model's key assumption regarding adverse selection, including differing levels of adverse selection by age. These analyses offer valuable insights into the long-term sustainability of PHI under varying conditions. The chapter further incorporates the impact of the COVID-19 pandemic by adjusting for changes in mortality, economic conditions, and claim experiences. Finally, it explores potential policy options designed to enhance the sustainability of the PHI market, with a particular focus on mitigating adverse selection through measures like age-differentiated premium increases.Item Open Access 2003: A Spatial Odyssey: Referring to Space in Makasai(2003) Brotherson, AnnaThis thesis provides an overview of how speakers of the East Timorese language Makasai (also known as Makasae) refer to space and spatial relationships. A summary of basic Makasai grammar is provided, followed by discussion and semantic analysis of space-denoting verbs, postpositions, directionals and demonstratives. The study is based on data collected by the author and other researchers at the Australian National University between July 2002 and November 2003.Item Open Access Applied Epidemiology- Evaluating the Impact of the Implementation of Fiji's Clinical Guidelines for Diagnosis and Management of Leptospirosis in the Central Division(2025) Faktaufon, DanielThesis Abstract for the Intention to Submit I commenced my Master of Philosophy (Master in Applied Epidemiology (MAE)) with my field placement at the Fiji Centre for Disease Control (Fiji CDC) in 2023, with the intention to complete by December of 2024. In this thesis I present the projects completed, my experience, and the completion of my core competencies during my placement. Chapter one provides an introduction to the thesis. Chapter two, describes my evaluation of a surveillance system. I evaluated the Early Warning Alert and Response Surveillance (EWARS) system for Suspected Leptospirosis Syndrome (SLS)in Fiji, from March 2022 to March 2023. SLS data from EWARS database and Laboratory data from the National Public Health Laboratory database were compared and analysed using Spearman's coefficient A strong correlation was shown to exist between SLS reports and laboratory-confirmed cases for the study period at National and in three divisions with correlation of 0.9. Participants in the survey were well-versed in EWARS and the recent changes. More training and feedback were highlighted by participants to further improve reporting. Chapter three describes my participation in an outbreak. The outbreak was a foodborne one in the Western Division. This outbreak included 433 people who had consumed prepared meals at a hotel and developed gastrointestinal symptoms within a matter of hours. As part of the investigating team, the 10 steps of outbreak investigation were followed We used relative risk (RR) analysis to identify exposure and outcomes. A RR of 353.3 indicated a common food source that had been consumed which resulted in the outbreak. Chapter four describes my data analysis. I conducted a retrospective analysis of Leptospirosis Microagglutination Test (MAT) results from 2018-2023.Laboratory data was obtained and a univariate analysis was conducted to identify the demographics details of tested samples. Multivariate analysis was then used to compare the different leptospirosis serovars by gender, age group, year and division. Leptospira Australis was found to be the most dominant serovar during the study period and was associated with reservoirs such as cattle and pigs. Chapter five describes my epidemiological research project. This was based on analysing the impact of Fiji's Clinical Management of Leptospirosis Guidelines 2016 and their implementation. An online survey was developed and delivered using REDCAP and focal group discussions were utilised using a Strengths Weaknesses Opportunities and Threats (SWOT) approach. The data were analysed thematically into five main areas. The results will contribute to the Fiji Ministry of Health review of the Leptospirosis guidelines and provide evidence-based recommendations to strengthen its implementation for healthcare workers. Chapters six and seven describe my teaching experiences and additional public health activities. These include introducing my MAE cohort to my field placement at Fiji CDC and describing how the public health systems in Fiji operate; the development of a research proposal for Japanese Encephalitis Virus preparedness Strategies in Fiji and teaching the 2024 first-year MAE cohort about Public Health systems at a regional level. Chapter 8 is a culmination of all work related experiences during my field placement, the lessons learnt and its application to my MAE program, including literature reviews, abstract writing, lessons from the field, conference and workshop presentations and my presentation at the Australian Communicable disease and Immunisation Conference (CDIC) in 2024.This work not only satisfies the competencies of the MAE program but will also continue to inform routine public health activities undertaken at Fiji CDC.Item Open Access The Dynamics of Social Identities in Postcolonial Hong Kong(2025) Ip, Wing ShanAims and Purpose: This thesis aims to improve the understanding of social identity evolution in Hong Kong, narrow the scope of plausible predictors of the dynamics of social identities, and be guided by the application of social identity, self-categorisation, and integrated threat theories to interpreting social identification in Hong Kong. There exists a large body of literature on intergroup conflicts arising from the tensions between national identity and ethnic or regional identity due to their undesirable social and political consequences. Hong Kong is potentially a valuable addition to this literature because it enjoys a fusion of Western and Eastern culture in light of its historical background as a British colony and the return of sovereignty to China in 1997. Over the past decade, the increasing awareness of localism in Hong Kong was manifested by a series of significant and unprecedented protests, specifically the Anti-Extradition Law Amendment Bill Movement (Anti-ELAB). Hong Kong Public Opinion Research Institute (HKPORI) conducted repeated cross-sectional surveys on Hong Kong residents' identification from 1997 to 2022. Previous researchers have examined the HKPORI data but they analysed only part of the datasets rather than the entire series. The stability of the relationships between identity strengths and their associated factors in the postcolonial era and how these might have changed if they were not stable remained unexplored. Method: This thesis is the first to use meta-analytical techniques to examine the dynamics of Hong Kong people's social identities between 1997 and 2022. In addition, mixed linear regressions and mixed logistic regressions were employed to test hypotheses while allowing for relationships between variables to change over time. Findings: The four empirical chapters in this thesis analysed trends and interrelations among six social identities (i.e., Hongkonger, Chinese, Citizen of People's Republic of China, Member of the Chinese Race, Asian, and Global Citizen) and four identity categories (i.e., Hong Kong Citizen, Chinese Citizen, Hong Kong Chinese Citizen, and Chinese Hong Kong Citizen). In particular, the compatibility between Hongkonger and Chinese identities as well as their predictors were investigated. The first empirical chapter provided an overview of the datasets and established the measurement properties of the variables by exploring trends and associations among key variables. The second empirical chapter extended Steinhardt et al. (2018)'s investigation of predictors of strength of identification as Chinese and Hongkonger. Over time, trust in the HKSAR government became a stronger predictor of Chinese identity strength but a weaker predictor of Hongkonger identity strength. It was found that the compatibility between Chinese and Hongkonger identities dropped drastically during the Anti-ELAB period. Anti-ELAB increased Hongkonger identity strength among those who identified as Hong Kong Citizen, Hong Kong Chinese Citizen, and Chinese Hong Kong Citizen, but reduced it among Chinese Citizen identifiers. The third empirical chapter extended Wong et al. (2021)'s examination of cohort effects on identities. While replicating their cohort effects by using their variable that forced respondents to choose between Hongkonger and Chinese identities, this study used separate identity strength measures and showed that the cohort effects occurred only for Chinese identity. The final empirical chapter tested a set of hypotheses developed from social identity theory, self-categorisation theory, and integrated threat theory. Confidence in "One Country, Two Systems", Hong Kong's future, and China's future positively predicted compatibility between Hongkonger and Chinese identities.Item Open Access Ecological assessment of the limestone habitat of Vietnam's Critically Endangered primates: Trachypithecus poliocephalus, Rhinopithecus avunculus, Nomascus nasutus, and implications for conservation(2025) Apthorp, KirrilyNon-human primates in Vietnam are facing increasing threats to their survival due to habitat destruction as a result of human expansion and development. As a biodiversity hotspot, Vietnam is home to a high density of non-human primate species, including some of the most Critically Endangered primates in the world. Three of these endangered primates - the cao vit gibbon (Nomascus nasutus), the Tonkin snub-nosed monkey (Rhinopithecus avunculus) and the Cat Ba langur (Trachypithecus poliocephalus) - are under serious threat due to habitat reduction and fragmentation. While occupying different ranges across northern Vietnam, these three species all inhabit limestone karst forest, an understudied and difficult to access habitat. Using a combination of in situ sampling and MaxEnt distribution analysis, this study compared the ecological profiles of each species and compared them to sites of potential habitat expansion to address the habitat fragmentation problem facing each species. Findings suggest that for the Cat Ba langur there may be viable corridor connections to increase connections between sub-populations, though suitability of this region for long-term settlement requires further investigation. Further work on overcoming potential corridor interruptions is also required. Results for the Tonkin snub-nosed monkey indicate that efforts to improve connections to the larger Du Gia is not recommended due to the high level of human activity is this region, including logging and mining. For the cao vit gibbon, results further support earlier recommendations by Fan et al. (2013) on the conservation prioritisation of an adjacent patch of forest as a potential habitat extension. Further recommendations for the cao vit gibbon include consideration of the introduction of an artificial corridor to supplement a discontinuous corridor between the two regions. Across all species, I recommend further long-term study of each area of potential expansion to further support conservation planning for these at-risk species. This research provides important information for conservation groups to plan for the future of these Critically Endangered species.Item Open Access Spectral Properties of Non-local Schr\"{o}dinger Operators: Techniques and a Counterexample(2025) Chen, SophieThis thesis aims to expand our understanding of the spectral properties of non-local Schr\"{o}dinger operators on an open set in Euclidean space. The first part concerns extending the method of modulus of continuity for solutions of parabolic equations---as used, for instance, to prove the Fundamental Gap Conjecture---to solutions of non-local heat equations on $\R^n$ and in dimension one with a non-local Neumann boundary condition. Specifically, we show that if a solution of a non-local heat equation has an initial modulus of continuity satisfying simple criteria, then this modulus of continuity is preserved at all subsequent times. In the process of trying to generalise our result in one dimension, we found a counterexample suggesting that a non-local analogue of the Payne-Weinberger inequality would depend on more than the diameter of a bounded (convex) domain. In the second part, we construct a counterexample demonstrating that the second eigenfunction of a perturbed fractional Laplace operator on a bounded interval, with Dirichlet `boundary' data off the interval, can exhibit more than one sign change. This stands in stark contrast to the classical expectation that it should have exactly one zero. Our construction employs the Kato-Rellich regular perturbation theory to analyse an infinite potential well eigenvalue problem, and then uses an energy-minimisation argument to extend this counterexample to finite potential wells. Although our detailed analysis focuses on the case $s=1/2$ (the Cauchy process), our approach strongly suggests that similar phenomena occur for other rational values of $s$ in $(0,1)$. At the time of writing, this result provides one of the first rigorous insights into the qualitative behaviour of eigenfunctions for perturbed non-local Schr\"{o}dinger operators.