Search

Deep Neural Networks (DNNs) are powerful computing systems that have revolutionised a wide range of research domains and have achieved remarkable success in various realworld applications over the past decade. Despite their significant recent advancements, training DNNs still remains a challenging task due to the non-convex and (potentially) non-smooth nature of the objective function. Back-propagation in combination with gradient-based minimisation approaches has been the predominant strategy for training DNNs for decades. Yet the popular error backpropagation algorithm is susceptible to potential drawbacks and limitations, for example its non-parallelisablity and biological implausibility, and vanishing or exploding gradients issues, etc. Inverting DNNs to infer likely inputs of the system from given outputs, is the other side of the same coin. Early ideas of DNNs inversion trace back to the 1990s, but research interests in more generic network inversion problems have been rekindled and primarily driven due to the rapid advancements in generative modelling in recent years. While several approaches for the inversion of DNNs have been proposed, the stability of the inversion is an often neglected crucial aspect. The neural network inversion problem is ill-posed as the solution does not depend continuously on the input datum hence can be highly sensitive to perturbations. The core theme of this thesis is at the training of DNNs. Built up on distributed optimisation approaches, this work contributes to both the learning problems and the inversion problems of DNNs. In particular, we propose a lifted Bregman learning framework that goes beyond the classical back-propagation approach, and aims to address unresolved and overlooked issues in training and the inversion of DNNs. More specifically, we propose a family of loss (penalty) functions that are based on a tailored Bregman distance. We provide detailed mathematical analysis on the derived Bregman learning framework and propose a whole range of deterministic and stochastic optimisation strategies to enable solving the learning problem. Bringing techniques and tools from Inverse Problems and Regularisation Theory, we provide theoretical guarantees as well as computational optimisation strategies for the stable, model-based inversion of neural networks.

Increasing research attention has been placed towards distributed dynamic systems that integrate sensing, computation, communication and physical processes. A prominent feature of such systems is communication among subsystems via network mediums, and distributed estimation and control schemes can enable distributed dynamic systems to carry out complicated tasks in a cooperative manner. Utilization of networks and exposure to the physical environment means that distributed systems are more vulnerable to adversaries that include attacks, faults, and disturbances. In practical distributed systems, a wide range of adversaries exist in various forms (such as deception attack and denial-of-service attack) that affect various channels (such as the sensor channel, communication channel and actuator channel). This thesis focuses on the important topic of developing resilient estimation and control schemes for distributed dynamic systems to improve the safety and performance in the presence of adversaries. Firstly, the simultaneous presence of disturbances and attacks on distributed systems is tackled by a novel three-stage estimation approach which includes anti-disturbance estimation, optimal attack detection and detection-triggered attack-resilient estimation. This approach effectively decouples the influence of multiple disturbances and false data injection attacks existing on the same channel. In some cases, heterogeneous attacks on different channels can be simultaneously injected to have a joint effect on distributed systems. Utilizing a novel event-based update scheme, an adaptive term and a distributed disturbance observer, an event-based distributed estimation approach is introduced to deal with the joint effects of aperiodic denial-of-service attacks on the communication channel and unknown deception attacks on the sensor channel. While the previous works deal with disturbances and attacks through compensation and attenuation, some attacks, namely the sparse injection attacks on sensors are potentially unbounded and cannot be dealt with the observer-based compensation approach. In such cases, it is more desirable to isolate and remove the sensor channels that are under attack. In the third section, a switching sparse attack detector based on a monitoring function utilizes the sensing redundancy to identify and remove the attacked sensor channels, and a backstepping control scheme is designed for the practical implementation of the proposed algorithm on a robotic manipulator. In the final section, adversaries on the actuator channel are studied, where the challenging topic of unknown direction faults on multi-input-multi-output distributed systems is dealt with novel Nussbaum functions, and a distributed containment control scheme is proposed for a network of uncertain nonlinear agents. Moreover, an event-triggering mechanism is introduced to avoid continuous communication among agents. The main contribution of this thesis is presenting a framework for the resilient estimation and control of distributed network systems against a wide range of adversaries, including but not limited to injection attacks, denial of service attacks, actuator faults and disturbances. The approaches introduced in each chapter of this thesis have compelling features that can be either implemented on their own, or integrated with other existing control and estimation schemes to enhance the resilience of distributed systems.

This practice-based research project sets out to adapt the literary concept of heteronym from Fernando Pessoa's fictional writing to performance practice to create three different fictional characters that reformulate the body. Being a Body and Becoming Somebody comprises fictional characters that combine the literary concept of the heteronym, performance practice, and the methodology of fictioning. Each fictional character gives rise to new subjectivities that re-engage with the original promise of the body - that of connecting between the self and the Other. It is an attempt to not only imagine reality using fictioning but also to express and produce new subjectivities in reality through the performance practice of fictioning. This practice-based research project sets out to adapt the literary concept of heteronym from Fernando Pessoa's fictional writing to performance practice to create three different fictional characters that reformulate the body. Being a Body and Becoming Somebody comprises fictional characters that combine the literary concept of the heteronym, performance practice, and the methodology of fictioning. Each fictional character gives rise to new subjectivities that re-engage with the original promise of the body - that of connecting between the self and the Other. It is an attempt to not only imagine reality using fictioning but also to express and produce new subjectivities in reality through the performance practice of fictioning. It aims to address the following research questions: What does it mean to use the body as an archive and an experimental testing ground for the multitude of different perspectives and fictional narratives? How does the process of becoming and fictioning affect and challenge the construction of my own identity as an artist? In addition to having Pessoa as my creative companion, I will also refer to Gilles Deleuze in search of a perpetual state of 'becoming' to complement the fleeting creation of multiple selves as they develop through performance. Finally, reflecting on the inescapability of being a body, I consider Gilles Deleuze and Félix Guattari and Rosi Bradotti's 'schizo-analysis' (Deleuze & Guattari, 1987) and 'difference machine' (Braidotti, 2011), to instrumentalise the three fictional characters in the iterative and diffractive practice of fictioning. Combining performance with reflective thinking, with an eye towards artistic creation, this project offers a hybrid practice-based enquiry into a performance-situated ontology that collapses the imaginative reality and possible futures and frames the contemporary notions of performativity, subjectivity and the role of the artist.

Traditional consumption-based asset pricing models generally treat the aggregate stock market as a claim to aggregate consumption. However, according to empirical evidence, the majority of households consume primary out of wages and live with no capital gains from risky investments, this implies potential omitted stock market systematic risk sources beyond aggregate consumption. Therefore, this thesis contributes to investigating alternative pricing instruments other than aggregate consumption growth to capture omitted stock market systematic risks in the traditional consumption-based capital asset pricing model (CAPM). This thesis constructs three asset pricing models for wealth redistribution risks, asset pricing effects of heterogeneous beliefs among wealth classes, and financial stress risks. A new market systematic risk dimension is uncovered in each model and then tested using U.S. stock market data. This thesis firstly incorporates the elevated stockholder consumption volatility into the long-run risk framework to account for wealth redistribution risks. The capital variability factor (CRV) is developed to capture the accumulated short-run volatility effects of wealth redistributive shocks. Secondly, this thesis derives the analytical solution of stockholder consumption growth to account for stockholder consumption risks using a robustness control approach. The stockholder consumption share factor (SCS) is constructed to portray the stockholder consumption dynamics and affords high asset pricing. Finally, this thesis utilises the long-run risk framework to factor in that financial stress contributes to the convexity of the price-dividend ratio and, therefore, serves systematic risks. The financial stress factor (FS) is developed to explain the cross-sectional return variations of stocks with highly subjective valuations and are difficult to arbitrage. This thesis contributes to the field in multiple dimensions. This thesis attempts to introduce economic variables into data-driven models, therefore shedding light on the impact of limited participation of households in the stock market for future research. Also, new market systematic risks are uncovered and counterpart asset pricing factors are constructed. These findings are important to policymakers as financial systems are especially vulnerable to systemic risk that cannot be hedged.

Creative practice is experiencing a revival as both research method and vehicle for impact and public engagement in geographical research. Academics are increasingly collaborating with artists to deliver public engagement and impact with an emphasis on the role of fine arts (e.g., Hawkins 2013a, 2013b, 2016). More grassroots-oriented participatory arts projects also tend to assert the benefits of art collaborations but often fail to acknowledge what happens in the space of artmaking. Recognising artists' roles in designing and delivering research challenges academics. To explore the gap between academics and artists, this project examines participatory art and artists' contributions to public education on plastics and the creation of public plastic literacy. Tracing the lessons learned through a series of public plastic junk-modelling workshops and exhibitions held in six different locations in the UK and the Philippines, the analysis draws on participant observations. This data is combined with autoethnography, site visits, interviews, questionnaires, and creative surveys undertaken with artists and workshop participants. Seeking the best methods and methodology to deliver participatory arts education on this messy but vital theme, the researcher's own journey remodels her into a creative geographer and a participatory arts facilitator/practitioner. The approach developed here generates a social typology of plastics and demonstrates participatory art's potential to deliver research while engaging publics and inspiring transformative behavioural change. This project thus fills the critical gap between celebrations of fine arts and the insistence on the value of an open-ended 'artfulness' (Pigott, 2020) that fails to fully consider artists in the research process. It demonstrates the intellectual and transformative power of research that fully incorporates artists' methodological contributions into its design.

Bioseparation is a major bottleneck in the manufacture of biopharmaceuticals such as proteins. Crystallisation is a cost-effective, rapid, and robust alternative technology to conventional chromatography steps in downstream separation processes. This thesis aims to investigate protein crystallisation as a feasible approach to separate proteins from a mixture for bioseparation. In this work, lysozyme- thaumatin mixture is used as the model binary protein mixture. Both µL-scale hanging-drop vapour-diffusion (HDVD) and mL-scale batch crystallisation methods were employed. We report an experimental evidence of direct selective protein crystallisation from a binary protein mixture solution where both proteins are supersaturated and crystallisable under an identical crystallisation condition. Results from both methods showed that protein impurity, even at low concentration level, would delay target protein crystallisation with an extended induction time. When silica particles were introduced as nanonucleants to facilitate crystallisation, target protein crystallisation was significantly improved with much shorter induction time. It was also indicated that the effectiveness of silica on crystallisation depended on the type of silica particle, silica loading amount, and impurity concentration. This study also revealed the critical role of agitation in obtaining consistent and reproducible results when moving from preliminary qualitative screenings using HDVD method to quantitative batch crystallisation experiments. Apart from improving reproducibility of crystallisation experiments, agitation also had impacts on both crystallisation rate, yield, and crystal size. In conclusion, this work demonstrates that protein crystallisation is a feasible and scalable methodology to separate a target protein from a complex mixture environment.

Analysing multiple types of omics data is a keystone methodology in biomedical research nowadays. With the significant advances of high-throughput experimental technologies, an enormous amount of omics data with extremely high dimensionality are generated every day at an unprecedented speed, which leaves a massive data gold ore waiting to be mined. However, it is difficult for traditional bioinformatics methods to deal with the high dimensionality and enormous data amount. The rapid development of machine learning, especially the deep learning methodology, has dramatically revolutionised fields like natural language processing and computer vision over the past decade. Deep learning has shown great success in decoding high-dimensional data like images, which makes it promising to adopt this cutting-edge technology to the analysis of multi-omics data. In this thesis, we propose a comprehensive toolbox for deep learning-based multi-omics data analysis named OmiSuite to decode high-dimensional multi-omics data and unveil the correlation between the phenotype profile and different types of omics profiles. OmiSuite is comprised of four components: OmiVAE, OmiEmbed, XOmiVAE, and OmiTrans. Among them, OmiVAE is one of the first endeavours to decode high-dimensional multi-omics data using variational autoencoders for pan-cancer classification. OmiEmbed is a unified multi-task deep learning framework for multi-omics data, supporting multi-omics integration, dimensionality reduction, omics embedding, tumour type classification, phenotypic feature reconstruction, survival prediction, and multi-task learning for aforementioned tasks. XOmiVAE is the explainable upgrade of OmiVAE, which can provide the contribution score of each molecular feature and each latent dimension for each phenotype prediction. OmiTrans is the first generative adversarial networks-based omics-to-omics translation framework, which ushered in a brand-new research topic with a promising vision. These four components of OmiSuite created a unified ecosystem for high-dimensional multi-omics analysis, which covered almost every aspect of this field, benefited follow-up studies and led to an upsurge of research in this field.

Childhood is a period of protracted gains in cognitive control, with increasingly flexible and efficient engagement of cognitive resources as a function of task demands. In particular, children progressively shift from relying on "as-needed" reactive control engagement at preschool age to more anticipatory proactive control engagement by middle childhood. Cognitive control is usually implemented in emotionally or motivationally charged contexts in real-world, which may present both challenges and opportunities for control engagement during childhood. The aim of this thesis was to examine how affective factors influence cognitive control in children, particularly the temporal dynamics of control engagement. Using electroencephalography (EEG), Study 1 showed that both positive stimuli and external rewards encouraged children and adults to engage practive control in a cued task-switching paradigm, but in different ways: provision of external rewards promoted proactive cue alertness, whereas the presence of positive stimuli enhanced proactive task selection. Study 2 leveraged fMRI to examine the brain circuitry supporting these beneficial effects in young adults. Provising of external rewards elicited greater cue-locked activation in the lateral frontal cortex, while no effect of positive stimuli on brain activation was detected. Study 3 used pupillometry to further investigate the effects of different reward types on control engagement. Both performance contingent and non-contingent rewards, provided as either a sustained incentive context or a transient incentive, improved proactive control engagement in the AX-CPT in childhood as well as adulthood. Willingness to engage cognitive effort, which was not influenced by reward contingency, mediated partially the effect of age on proactive control engagement. Finally, Study 4 compared the effects of social and non-social rewards on cognitive control engagement in an online version of AX-CPT, with other important reward dimensions being balanced between reward types. Social rewards enhanced cognitive control performance but did not influence whether control was engaged reactively or proactively in both preschool and school-aged children, whereas no effect of non-social rewards was observed. Together, these findings indicate that external rewards and, to a lesser extent, positive stimuli, can help children strategically upregulate cognitive control engagement by promoting a proactive mode of control, although these effects do not always translate into behavioural benefits. They have important implications for assessing and promoting cognitive control both in and outside of the lab.

The More-Electric Aircraft (MEA) concept is a major trend in aircraft electrical power system engineering and results in an increase in onboard electrical loads and electrical power demand. On an aircraft the jet engine acts as the prime power source, providing to the onboard systems pneumatic power, hydraulic power, mechanical power and electrical power. As the aerospace industry is moving towards the MEA paradigm, many functions which are conventionally driven by pneumatic, hydraulic and mechanical power systems are being replaced by their electrical counterparts. However, due to the power off-take limit from the high-pressure (HP) spool, extra power is considered to be extracted from the low-pressure (LP) spool of an engine through the coupled generators. Increased number of generators and power electronic penetration renders a higher requirement to the power generation centre (PGC) in a more electric engine in terms of controllability, stability, fault tolerance, and power flow flexibility. In recent years, a dual-channel single DC bus PGC was developed for the MEA application. Two electrical machines are coupled to the HP and LP spools of the engine respectively, and each of them is independently regulated by an AC/DC power converter. The DC outputs of the converters are parallel connected to supply a common DC bus. Theoretical analysis and experimental results of relevant research confirm the superiority of the PGC in efficiency and power sharing between sources. However, the PGC suffers from three critical drawbacks: 1) poor fault tolerance to the main rectifier failure; 2) power losses caused by the significant de-fluxing current in the HP generator at high-speed settings, and 3) the incapability to achieve engine and HVDC grid stability optimization in the high-power settings of the engine. To address these drawbacks, an advanced power generation centre (APGC) with an AC/AC bridging converter is proposed in this Thesis. The bridging converter is introduced to make the APGC much more flexible than the PGC in terms of managing the power flow and the generators control. Compared with the PGC, the APGC allows post-fault operation and supplies uninterruptible power to the downstream loads in the case of main rectifier failure. It also allows the permanent magnet synchronous machine-based HP generator to operate at a high speed without field-weakening operation, which significantly reduces the stator currents and power losses in the HP generator and power converters. Moreover, due to establishing a power exchange path between the HP and LP power generation channels, the stability of the onboard HVDC grid and the engine can be enhanced simultaneously in the high-power settings of the engine. System configuration, control schemes in normal and post-fault operation modes, power losses and stability with different power sharing ratios between sources of the APGC have been thoroughly analysed in this Thesis. Theoretical findings of the APGC support the above-mentioned improvements over the PGC. The technical results and simulations are supported by Matlab/Simulink based models and validated by experimental work on an engine emulator system and a downscaled lab prototype of the APGC.

Numerous bacteria associated with human health (probiotics and pathogens) exhibit inherent properties in terms of targeting, penetrating, proliferating and killing solid tumours through various mechanisms. To rationally redesign these microorganisms into living medicine, there is a lack of genetic software with suitable induction characteristics and compatible cellular chassis with enhanced biosafety, owing to the stringent requirements and controllability during medical practice. Consequently, the development of bacteria therapy lags behind compared to other applications offered by synthetic biology. The goal of this thesis is to address these two significant challenges to realise the full potential of bacteria therapy. Three different transcriptional control systems were constructed: (1) an aspirin/salicylate responsive regulatory SalR:PSal system (2) a DNA damage responsive Vibrio recA SOS system and (3) a nitric oxide responsive NorR:PNorV system. The system performance was rationally re-designed and optimised based on the detailed molecular mechanism. A universal competitive binding mechanism was found during the tuning process, where the expression level of the regulator played a dominant role in system dynamics and displayed dual activator/repressor features dependent on the state of inducer binding. To enhance the biosafety and usability, these genetic software were characterised and optimised in the cellular context of a probiotic chassis strain, E. coli Nissle 1917 and tested in the non-dividing anucleate E. coli mini SimCells, which are minicells with intact transcriptional/translational machinery. An optimised purification protocol was developed with a final yield of 5 × 10^9 mini SimCells/ml and high purity (total elimination of parent cells). Optimised sensor construct loaded in mini SimCells displayed functional responsive dynamics that (1) are robust against environmental fluctuation, (2) displayed faster response time (<30 mins), and (3) maintained medically relevant sensitivity. These findingsvalidated the programmability of mini SimCells to execute both ơ54 and ơ70 based transcriptional circuits and bypass host interference caused by global resource allocation or external fluctuation. To conclude, this thesis provides a systematic workflow of designing therapy tailored transcriptional software and its functional deployment in mini SimCells. The circuits allow both external controls using established medical methods such as chemo/radiotherapy, aspirin injection or internal triggering via in-situ disease biomarkers, pave ways for the future pre-clinical development. Furthermore, this thesis demonstrates the potential of mini SimCells and its design guideline as a minimum version of bacteria-bot systems to achieve robust and sensitive performance, making it a preferred choice in clinical practice.

Orthogonal Frequency-Division Multiplexing (OFDM) has compelling benefit in Optical Wireless Communications (OWC). Hence, we analyse the Layered Asymmetrically Clipped Optical OFDM (LACO-OFDM), which is one of the most advanced Optical OFDM (O-OFDM) schemes, and improve its overall system performance. We commence in Chapter 1 with a review of the development of OWC and O-OFDM, discussing its Spectral Efficiency (SE) and Power Efficiency (PE). Then, LACO-OFDM introduced for striking a trade-off between the attainable SE and PE. This is followed by a literature review of its recent development. A more detailed tutorial on LACO-OFDM is then given in Chapter 2, relying on detailed examples. To further study the properties of LACO-OFDM, Chapter 3 analyses its 1) Probability Density Function (PDF) and its statistical moments; 2) Bit Error Ratio (BER); and 3) Peak-to-Average Power Ratio (PAPR). As a further advance, we propose a toneinjection aided PAPR reduction design for LACO-OFDM, which in turn improves the BER performance. Simulations are provided for verifying both the analytical BER performance and the PAPR distribution of LACO-OFDM. The results show that the expressions derived match well with the simulations. Furthermore, the PAPR reduction method proposed attains a 5 dB PAPR reduction at the 10−3 probability-point of the complementary cumulative distribution function, as well as a better BER performance than the original LACO-OFDM scheme. For the sake of improving the BER performance, multi-class channel coded LACOOFDM is proposed in Chapter 4, where the channel capacity of the system is derived based on our mutual information analysis. We conceive a multi-class channel encoding scheme integrated with the layered transmitter. At the receiver, both the coded and uncoded likelihood ratios (LLRs) are extracted for inter-layer interference cancellation and symbol detection, respectively. Simulations are conducted and the results show that our design approaches the achievable rate within 1.1 dB for 16QAM 4-layer LACOOFDM with the aid of a half-rate 8-iteration turbo code at BER = 10−3 , outperforming its conventional counterpart by about 3.6 dB. iv In Chapter 5, we propose algorithms for optimising the Discrete-input Continuousoutput Memoryless Channel (DCMC) capacity of LACO-OFDM. Then, an algorithm is proposed for maximising the capacity for twin-layer LACO-OFDM by optimising the power sharing between the layers. This is followed by the conception of a more general algorithm applicable to LACO-OFDM having an arbitrary number of layers. Numerical results are provided for quantifying the capacity improvement attained by the proposed algorithm. Moreover, an adaptive layer-activation scheme is proposed for adjusting the number of layers to be used for maximising the capacity at different SNRs.

At present there are already many deep geothermal projects allowing us to have a better understanding of deep geothermal energy. However, there are still many issues to be solved for more reliable use of deep geothermal energy. This thesis proposes an engineering analysis method to assess the performance of a typical deep geothermal system, which is a doublet system with one injection well and one extraction well. A convective heat transfer boundary between the aquifer and overburden layers is applied for the axisymmetric problem based on the Lauwerier model for the first time. A new analytical solution is deduced using a series of Laplace transforms. The interaction between the injection well and extraction well is first neglected. Compared with other relevant analytical solutions, this new analytical solution comprehensively includes both heat conduction and heat advection in the aquifer and the heat flux between the aquifer and overburden layer. As long as the relevant parameters are obtained, the new analytical solution can intuitively illustrate the temporal and spatial temperature distribution within the aquifer. It can be used to determine the location of the extraction well and to evaluate the extracted geothermal power of the hot water aquifer. The convective heat transfer boundary at the interface does not only reflect the actual heat transfer process at the interface, but also models the heat transfer process in the overburden layer. It is shown that the dimensionless equivalent heat transfer coefficient is expressed as a function of the dimensionless injection rate and the dimensionless thermal conductivity of the overburden layer so the new analytical solution effectively incorporates the properties of the overburden layer. A series of FE simulations are conducted, and the analytical model is curve fitted to the FE results to evaluate the values of the dimensionless equivalent heat transfer coefficient. Based on the results of the curve fitting exercise, two empirical equations are proposed for typical cases. Applying the analytical solution coupled with these empirical equations and along with proper error estimates, it is possible to conduct a simple and rapid evaluation of the geothermal potential of a particular site. The revised analytical solution in this thesis is novel as there is no other analytical or semi analytical solution for the doublet system considering the heat conduction and heat advection in the aquifer and the heat flux between the aquifer and overburden layer. The revised analytical solution extended the analytical solution for a single injection well to a doublet scheme by considering the interaction effect between the injection well and the extraction well. The expression of the critical distance between two wells is obtained so that the best location of the extraction well in a doublet system can be determined. The spatial and temporal temperature distribution in the aquifer for a doublet scheme can be given by the revised analytical solution when the well distance is greater than the critical distance. It is found that it is valid to use a single well model to simplify a doublet scheme when the extraction well is far away from the injection well. The temperature of the extracted water against different time, injection rates and well distances can be obtained via the revised analytical solution. The revised analytical solution is compared with the experimental data and the numerical solutions and it is found that they match with each other well. The effect of a natural fault/fracture that exists in the aquifer on the performance of a doublet system, namely the temperature distribution and extracted temperature, is evaluated. By comparing the line model with the domain model, it is found that the line model is valid and computationally efficient. It is found that the acceleration effect of the fracture on thermal movement is the greatest when the fracture is located at the midpoint of the two wells. When the fracture is shifted towards the injection (extraction) well, the acceleration effect decreases and then becomes the deceleration effect. The deceleration effect of the fracture is the greatest when the fracture is located at the injection (extraction) well. The expressions of the critical angle for any position of the fracture in the doublet system are obtained. Equipped with these expressions, it is possible to decide whether the doublet system is still efficient during its life span once the cold water injection rate and the geometry and properties of the fracture are given.

The aim of this thesis was to provide a comprehensive examination of the roles of macrofauna in estuarine biogeochemistry by adopting a multidisciplinary field and laboratory-based approach and integrating species distribution modelling with biomass-dependent ecosystem function relationships. Eventually, an ecosystem management tool to provide insights into the ecological consequences of anthropogenic disturbance in estuaries was developed using the Western Scheldt as a model system. To establish a baseline understanding of the spatio-temporal patterns of macrobenthos (activities) and ecosystem functioning in the Scheldt estuary, a seasonal survey was carried out during four consecutive seasons in 2015 and 2016, sampling three habitats with different hydrodynamic regimes (low-dynamic intertidal, high-dynamic intertidal, subtidal) in each of the three main salinity zones (polyhaline, mesohaline, oligohaline). Fluxes of dissolved oxygen, nitrate, nitrite, ammonium and phosphate were measured, as well as environmental properties of the sediment and water, macrofaunal activity (bioturbation and bio-irrigation rates) and macrofauna trait composition with respect to sediment reworking. Luminophores were applied on top of the sediment to measure sediment particle mixing modes and rates. The decline in the concentration of the added inert tracer sodium bromide in the water column was used to estimate pore water exchange rate as a proxy for bio-irrigation. Data and insights obtained from this field survey are presented in Chapter 2 and Chapter 3. Chapter 2 assesses the spatial and temporal variability of these benthic ecosystem processes (i.e. particle mixing and bio-irrigation) in the Scheldt estuary. Luminophore profiles showed that biodiffusion was the dominant particle reworking mode. Rates of both particle reworking and bio-irrigation were highly variable among seasons and habitats, and different species were found to contribute unevenly to both processes, and between habitats and seasons. Habitat structuring effects on populations, density-dependent interactions with the habitat, and temperature-driven variability in macrobenthos activity and living position in the sediment are suggested to explain the observed spatial and temporal differences. In Chapter 3, the relative contributions of macrofauna and the environment to the spatial and temporal variability in benthic biogeochemistry were investigated. Using distance-based redundancy analysis, we found total density, bio-irrigation and temperature to be the main contributors to biogeochemical variability, but this model explained only 23 % of the total variation. Variation partitioning and analyses of subsets of the data in the different seasons, showed that processes linked to the biota were lowest and overruled by environmental steering in the coldest period of the year, while processes related the burrowing behaviour of macrofauna, or densities and biomasses of specific sediment reworking functional groups, predominantly determined biogeochemical variability from June onwards. Macrofaunal contribution to biogeochemistry was highest in low-dynamic intertidal habitats that are densely inhabited by biodiffusing and bio-irrigating fauna. We conclude that the impacts of macrofaunal and environmental factors on the biogeochemical fluxes in the Scheldt estuary vary along the estuarine gradients and with season and are thus highly context-dependent. This spatial and temporal variability should therefore be considered to extrapolate biogeochemical fluxes to entire ecosystems. The impact of two bioturbating benthic invertebrates, Limecola balthica and Hediste diversicolor, on sediment community oxygen uptake in the Scheldt Estuary was examined in Chapter 4. Both H. diversicolor and L. balthica irrigate the sediment, the former by ventilating their burrows and the latter by siphoning water. Laboratory-controlled microcosms containing defaunated sediment amended with artificially composed faunal densities of different body sizes were used to test how species identity, habitat, and population density influence O2 uptake in different habitats (muddy and sandy sediments) in monoculture. Both L. balthica and H. diversicolor facilitated O2 fluxes between the sediment and the overlaying water, and a major portion of the variance in sediment metabolism and bio-irrigation could be explained by the per capita body size and density, or by the total biomass of the inhabiting bioturbators. H. diversicolor showed a more pronounced relationship between biomass/density and faunal-mediated O2 consumption than L. balthica. Respiration was significant in predicting faunal-mediated O2 uptake among different species (H. diversicolor and L. balthica) across different habitats (sandy and muddy sediments); whilst the relationship between bio-irrigation and faunal-mediated O2 uptake was significant for H. diversicolor in both sediment types and and for L. balthica in sandy sediments only, with a lower predictive power of bio-irrigation compared to respiration as the predictor. Analysis of covariance demonstrated significant habitat effects in biomass-dependent bio-irrigation, which might be attributed to different physical constraints (e.g. O2 availability) in sandy and muddy sediments. In summary, we demonstrated that the faunal-mediated O2 consumption by macrobenthos is density- and biomass-dependent, but the extent to which the variance of faunal-mediated O2 uptake can be explained is also conditioned by the interplay between the abiotic environment and the biological traits of the species, which is in support of our hypothesis that biological traits (e.g. bio-irrigation) and sediment physico-chemical properties significantly affect faunal-mediated O2 consumption. The specific example of using macrobenthos for ecology conservation and estuarine ecosystem management is explored in Chapter 5, with H. diversicolor as the model species, since this species was found to be the dominant contributor to biogeochemical fluxes in the Scheldt estuary (Chapter 2-3). We quantified the contribution of H. diversicolor across a range of biomasses in sediment metabolism and extrapolated the spatial variability of the faunal-mediated O2 consumption based on its biomass and distribution within its natural habitats in the Western Scheldt. Biomass-scaling of faunal-mediated O2 uptake by H. diversicolor was quantified from laboratory-controlled microcosms containing defaunated sediment with artificially composed faunal densities and body sizes in its habitats (polyhaline sandy sediment, polyhaline muddy sediment and mesohaline muddy sediment) along the estuarine gradients of the Western Scheldt; and the spatial variability of the fauna-mediated O2 consumption was extrapolated at the landscape scale by combining spatial mapping of H. diversicolor developed from multi-quantile regression modelling. Furthermore, the years 1955 and 2010 were compared to investigate changes in H. diversicolor-mediated O2 consumption in relation to the anthropogenic modifications of the estuary between both years. By tuning the quantile of the responses (upper quantile and full quantile), two species distribution scenarios were developed to describe organisms’ responses to different environmental constraints. Biomass of H. diversicolor was a highly significant (R2 = 0.9) determinant of fauna-mediated O2 uptake (F1,26 = 119.6, p < 0.001) that was unaffected by habitat type (F2, 26 = 3.12, p = 0.06). Therefore, the governing function of biomass scaling can be used to scale the biomass-dependent fauna-mediated O2 uptake to each of three different habitat types. In the intertidal region, maximal fauna-mediated O2 uptake estimated from the upper quantile regression (tau=0.95) was 8772 and 6201 mol d-1 in 1955 and 2010, respectively, corresponding to respective total biomasses of 4407.12 kg and 2915.03 kg H. diversicolor for the entire Western Scheldt.

In general, asteroids are of great significance for the future. This point has been reflected in several different fields, including scientific research, long-term economic development and the long-term global safety. Driven by these issues, the exploration and exploitation of asteroids has drawn significant attention and led to a series of space missions. Moreover, among the entire asteroid population, binary-asteroid systems represent a significant subset, which can be potential candidates for future missions. The purpose of this thesis is to develop the methodologies necessary for future binaryasteroid exploration and utilisation consists of two main objectives: the orbit manipulation of a binary-asteroid pair and capture mechanics for a binary system. In this thesis, a study of the low-thrust orbit manipulation of a binary-asteroid pair is presented, under the influence of the Sun's third-body perturbation. The Sun-binary hierarchical three-body system is represented as two interacting perturbed Keplerian motions: inner motion within the binary and outer motion around the Sun. Moreover, the third-body perturbations are simplified by the quadrupole approximation, and a first-order analytical approach is derived to evaluate the secular variation of the orbital elements for the system. Based on the analytical solution provided, open-loop control laws are then designed to obtain a desired change in the binary orbital elements. In addition, this thesis also investigates the feasibility of capturing one member of a binary asteroid pair through disruption of the binary asteroid during the flyby of a central body, e.g. the Earth. For simplification, the process is approximated by the planar parabolic restricted three-body problem (PPRTBP) and a set of related regularised equations of motion are derived. Based on the PPRTBP, a linear analysis is also conducted for motion in the vicinity of the main asteroid, which provides insights for the inner motion of the binary system. Moreover, initial condition maps for the evolution of the binary-asteroid pair determine the capture region in the initial configuration space. Two strategies have then been proposed to investigate the possibility of engineering the capture process. One is a re-phasing manoeuvre before encounter to engineer the initial condition maps, the other is an optimal single-impulse transfer during encounter, both of which have been verified by related numerical simulations.

This thesis reports the synthesis and characterisation of a series of layered oxyarsenides A2M3As2O2 (A=Sr, Ba and M = Cr, Mn) and layered oxychalcogenides Sr2MO3CuX (M = Cr, Mn, Fe and X= S, Se). These materials are synthesised with different choices of the transition metal cations and it has been demonstrated that the nature of the metals can strongly influence the magnetic properties of these layered solids. Variable temperature powder neutron diffraction on Ba2CrO2Cr2As2 has revealed long range antiferromagnetic ordering of the Cr2+ ions in both oxide and arsenide layers below 248 K and 470 K and the magnetic moments in both layers are aligned along the c-axis, where the two independent Cr2+ sublattices order with two different propagation vectors k = (½ ½ 0) in the oxide layer and k = (1 1 1) in the arsenide layer. The system has similarities and differences with the Sr analogue. The partial substitution of Cr2+ by Mn2+ in Sr2CrO2Cr2As2 and Ba2CrO2Cr2As2 has been shown to change their crystal structures and magnetic properties depending on the Mn:Cr ratio. Variable temperature neutron powder diffraction enabled the magnetic structures to be determined and the compounds with substitution all show antiferromagnetic ordering, analogous to those seen in their parent samples. The most characteristic feature of the magnetic properties in these compounds is the appearance of two similar magnetic structures, with the magnetic spins in adjacent arsenide (oxide) layers aligned either parallel or antiparallel. Similar to Sr2CrO2Cr2As2, a spin-reorientation of the moments is found in the arsenide layer of Sr2Cr0.93Mn0.07O2Mn0.92Cr1.08. PND measurements have revealed long-range AFM ordering in the oxide layers of Sr2MnO3CuSe. In replacing Se2- by S2-, a new magnetic structure is found in Sr2MnO3CuS which is different in detail to that of the selenide analogue. The deintercalation of a small amount of Cu in Sr2MnO3CuS results in adoption of the same magnetic structure as found for the selenide analogue and leads to a much larger ordered moment (3.60(4) μB/Mn3+) than in stoichiometric Sr2MnO3CuS (2.83(6) μB/Mn3+), with a value approaching that of Sr2MnO3CuSe (3.76(2) μB/Mn3+). The different magnetic behaviour between Sr2MnO3CuSe and Sr2MnO3CuS suggests that the Mn3+ ions in sulfide are slightly reduced and this introduces some Mn2+ ions and leads to moment reduction due to frustration. The neighbouring moments in all cases are found to be antiferromagnetically coupled in a checkerboard fashion with their moments directed perpendicular to the ab-plane. Investigations into Sr2FeO3CuSe have revealed a spin reorientation transition of Fe moments between 1.5 K and 230 K with the tilted Fe3+ spins in the oxide planes gradually reorienting to lie in the ab plane. Similarly, Sr2CrO3CuSe demonstrates antiferromagnetic ordering of Cr3+ in the oxide layers at 4 K with their moments aligned within the ab-plane. Finally, the synthesis of a novel and challenging compound Sr2CrO3CrAs has been successful. Variable temperature powder neutron diffraction reveals long range antiferromagnetic ordering of the Cr2+ ions in the arsenide layer below 478 K, with their moments directed along the c-axis. In contrast to other Sr2CuGaO3S-type compounds, there is no evidence for long range magnetic ordering in the oxide layers of Sr2CrO3CrAs.

Bayesian machine learning has gained tremendous attention in the machine learning community over the past few years. Bayesian methods offer a coherent reasoning for quantifying uncertainties in the decision making procedure, based on the Bayes rule. One of the core advantages of Bayesian methods is the separation of modelling and inference. In other words, the likelihood models are completely independent of the computation of the posterior distribution of the parameters. There are many Bayesian models that are widely used in the machine learning community. For example, non-parametric models such as Gaussian Processes and Dirichlet Processes are flexible models which are able to capture and learn the structure of the data. Bayesian deep learning models, which are based on neural networks, are another example of flexible Bayesian models that are rich enough to represent non-linear structures in the data. The process of inferring the posterior lies at the center of Bayesian inference. When computing the posterior distribution exactly is not feasible, due to intractability of the posterior and the computational or memory constraints, approximate Bayesian inference comes to play. In this PhD thesis, I develop and investigate various Bayesian modelling and inference techniques and apply them to multiple interesting domains and tasks. We begin with Tucker Gaussian Processes(TGP), a class of flexible non-parametric models based on Gaussian Processes (GP). We apply the method to 1) regression problems on structured input data, and 2) collaborative filtering problems where TGP offers an elegant way of incorporating side information. We demonstrate superior results compared with benchmarks on a number of examples across different domains. A closely related line of research based on GPs is Bayesian Optimization (BO). It is a black-box optimizer where one optimizes an objective function through subsequent queries about next input locations to be evaluated at. However, this method does not work well when the input space is non-Euclidean or combinatorial. We alleviate the problem by learning a low dimensional Euclidean representation of the combinatorial input space with variational inference, using Variational Auto-encoder (VAE). The optimization can then be conducted on the low dimensional embedding instead. We apply our method to Automatic Statistician and natural scene understanding, which give promising results. For approximate Bayesian inference, we first propose an algorithm called Relativistic Hamiltonian Monte Carlo (RHMC) which is a variant of MCMC. In particular, we replace Newton's kinetic energy in the Hamiltonian with Einstein's relativistic kinetic energy, which makes the algorithm more robust. There are several extensions to RHMC, including a stochastic gradient version for scalability, a thermostat version based on the temperature of the physical system and a resulting optimization algorithm which gives comparable performance compared with the state-of-the-art. Finally, we propose another sampling based inference method called the Adaptive Importance Sampling with Exploration and Exploitation (Daisee), where we look into the problem of exploration-exploitation in adaptive importance sampling through establishing a natural connection between importance sampling and multi-armed bandit problem. In particular, through a finite-time regret analysis we show that the regret of the proposed algorithm grows sublinearly with time. Further, we propose a hierarchical extension of Daisee to encourage exploration in the region with high uncertainty. The new models proposed in this thesis help to allow for more flexible Bayesian modelling and the inference techniques introduced can open new research directions for efficient and accurate posterior inference. These contribute to Bayesian inference and probabilistic machine learning.

This dissertation examines the practice of storytelling in norm diffusion based on an in-depth political ethnography of the United Nations Development Programme (UNDP) in China. Closely engaging with glocalisation literature, this dissertation focuses on the manner in which non-linear and multidirectional norm diffusion that disrupts, fractures and transcends the local-global opposition, creating entanglements across layers of orders and scripts. Norm translators are the main actors in the mediation and communication of reframing and relocating normative packages. Empirically, this dissertation addresses the insufficient attention paid to the everyday micro-dynamics of norm diffusion, and particularly how norm circulation and contestation unfold and intensify in the context of emerging powers. The UNDP China office thus serves as an ideal case to study how such practices and processes unfold over time. My methodological approach is political ethnography, employed through a period of intensive fieldwork and participant observation carried out from 2015 to 2018, when my work at the UNDP allowed unparalleled access to materials and the internal organisational processes. My fieldwork experience highlights the critical role played by stories in translating and appropriating norms. Drawing on the theories of in-between storytelling in institutional contexts, this dissertation argues that an internalisation process of "norm metamorphosis" translates original norms into textual realities, projects and stories. These stories are normative, canonical, ambiguous, open to interpretations, and used as persuasion instruments. Detailed case studies then demonstrate the emergence and application of stories in human rights, rule of law and development cooperation. The content, form and audience of the stories are shaped by issue characteristics and ideological backgrounds, bearing the hallmarks of UNDP officials and local actors. The dissertation further describes how local norms return to the global stage through the UNDP's narration of China's development experience. Instead of reinforcing the local-global opposition, these stories carve out intermediary and liminal spaces through which concepts and ideas travel and transform, redefining the identity of international organisations and transnational networks in recursive conversations.

Many scholars have addressed the important role that guanxi plays in the Chinese social-economic system. Guanxi is widely accepted in academia as an indigenous construct from China-deeply rooted in Chinese culture and reflected in the behaviour of Chinese people in both the social and institutional environment. During the growing globalisation of Chinese multinational companies, Chinese expatriates have increasingly taken up international assignments and inevitably carried Chinese guanxi to the host country. Research on guanxi in China has been intense. However, how the employees of Chinese multinational companies employ guanxi in the West, how Chinese expatriates develop and use guanxi in the host country, and how these behaviours affect their adjustment remain unclear. My dissertation contributes to this line of study in three ways. I first examined guanxi capitalism theoretically to highlight the features of China's economic system embedded deeply in its cultural-social-political environment and to explore how guanxi emerged, evolved, and subsequently dominated the economic system in China. Based on this analysis, I developed a conceptual framework of the "Guanxi capitalism structure" to illustrate the fundamental role of guanxi as the "invisible hand" in China. Next, I examined empirically how guanxi practice affects intra-firm multicultural group dynamics involving Chinese expatriates, host-country nationals, and host-country Chinese in Chinese multinationals. My study shows how expatriates actively practice guanxi with their homeland counterparts, but they do not do so with host-country nationals and host-country Chinese, and it explores the implications of these dynamics. Finally, I examined the impact of guanxi building on Chinese expatriate adjustment. Based on my analysis, I developed a process model that illuminates that guanxi development alters expatriates' adjustment curve significantly. My findings contribute more generally to shed light on cross-cultural management in terms of Chinese guanxi practice and the process of initiating, building, and utilizing guanxi in the Western context.

Language learning strategy (LLS) has been long considered as one of the most important variables impacting language performance. Learning context, as a major effect on LLS choices, has not been paid much attention compared to other individual variables in LLS research. For the purpose of highlighting contextual influence in LLS research, the present study investigated LLSs used by higher level students in a target language environment and an EFL/foreign language environment. To achieve the goals, a total of 44 enrolled doctoral students (19 in the United States and 25 in China) were recruited as participants. A background survey, a questionnaire (i.e., Language Strategy Use Inventory [LSUI]), and semi-structured interviews were deployed to gather data. The findings first showed that higher level students used a wide repertoire of LLSs to facilitate their language learning in each learning environment. Second, the findings identified eight specific contextual factors influencing the LLS use of higher level students in both learning environments, which included access to target language, participation in an academic environment, access to target language learning sources, experience as a teaching associate/research assistant, exposure to diverse accents and dialects of the target language, lack of feedback on errors, experience in target language classes, and beliefs about social reaction to learning target language. Third, the findings also showed that students reported some specific LLSs to be most useful in each learning environment. For example, "talking with (native) English speakers" was considered as a useful speaking strategy in target language environment, and "using language learning applications (APPs)" was considered as a useful listening and speaking strategy in EFL/foreign language environment. In addition, implications for future research and pedagogy were offered. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]

Distribution Automation (DA) is deployed to reduce outages and to rapidly reconnect customers following network faults. Recent developments in DA equipment have enabled the logging of load and fault event data, referred to as pick-up activity. This pick-up activity provides a picture of the underlying circuit activity occurring between successive DA operations over a period of time and has the potential to be accessed remotely for off-line or on-line analysis. The application of data analytics and automated analysis of this data supports reactive fault management and post fault investigation into anomalous network behavior. It also supports predictive capabilities that identify when potential network faults are evolving and offers the opportunity to take action in advance in order to mitigate any outages. This thesis details the design of a novel decision support system to achieve automatic fault diagnosis and prognosis for DA schemes. It combines detailed data from a specific DA device with SCADA data, by utilising rule-based, data science techniques (e.g. data mining and clustering techniques) to deliver the diagnostic and prognostic functions. These are applied to 11kV distribution network data captured from Pole Mounted Auto-Reclosers (PMARs) as provided by a leading UK network operator. This novel automated analysis system diagnoses the condition of device faults, the nature of a circuit's previous fault activity, identifies underlying anomalous circuit activity, and highlights indications of problematic events gradually evolving into a full scale circuit fault using prognostic functionality. The novel contributions also include the characterisation and identification of semi-permanent faults and a re-usable methodology and approach for applying data analytics to any DA device data sets in order to provide diagnostic decisions and mitigate potential fault scenarios.

Social signal processing aims to automatically understand and interpret social signals (e.g. facial expressions and prosody) generated during human-human and human-machine interactions. Automatic interpretation of social signals involves two fundamentally important aspects: feature extraction and machine learning. So far, machine learning approaches applied to social signal processing have mainly focused on parametric approaches (e.g. linear regression) or non-parametric models such as support vector machine (SVM). However, these approaches fall short of taking into account any uncertainty as a result of model misspecification or lack interpretability for analyses of scenarios in social signal processing. Consequently, they are less able to understand and interpret human behaviours effectively. Gaussian processes (GPs), that have gained popularity in data analysis, offer a solution to these limitations through their attractive properties: being non-parametric enables them to flexibly model data and being probabilistic makes them capable of quantifying uncertainty. In addition, a proper parametrisation in the covariance function makes it possible to gain insights into the application under study. However, these appealing properties of GP models hinge on an accurate characterisation of the posterior distribution with respect to the covariance parameters. This is normally done by means of standard MCMC algorithms, which require repeated expensive calculations involving the marginal likelihood. Motivated by the desire to avoid the inefficiencies of MCMC algorithms rejecting a considerable number of expensive proposals, this thesis has developed an alternative inference framework based on adaptive multiple importance sampling (AMIS). In particular, this thesis studies the application of AMIS for Gaussian processes in the case of a Gaussian likelihood, and proposes a novel pseudo-marginal-based AMIS (PM-AMIS) algorithm for non-Gaussian likelihoods, where the marginal likelihood is unbiasedly estimated. Experiments on benchmark data sets show that the proposed framework outperforms the MCMC-based inference of GP covariance parameters in a wide range of scenarios. The PM-AMIS classifier - based on Gaussian processes with a newly designed group-automatic relevance determination (G-ARD) kernel - has been applied to predict whether a Flickr user is perceived to be above the median or not with respect to each of the Big-Five personality traits. The results show that, apart from the high prediction accuracies achieved (up to 79% depending on the trait), the parameters of the G-ARD kernel allow the identification of the groups of features that better account for the classification outcome and provide indications about cultural effects through their weight differences. Therefore, this demonstrates the value of the proposed non-parametric probabilistic framework for social signal processing. Feature extraction in signal processing is dominated by various methods based on short time Fourier transform (STFT). Recently, Hilbert spectral analysis (HSA), a new representation of signal which is fundamentally different from STFT has been proposed. This thesis is also the first attempt to investigate the extraction of features from this newly proposed HSA and its application in social signal processing. The experimental results reveal that, using features extracted from the Hilbert spectrum of voice data of female speakers, the prediction accuracy can be achieved by up to 81% when predicting their Big-Five personality traits, and hence show that HSA can work as an effective alternative to STFT for feature extraction in social signal processing.

This thesis consists of three essays which utilize experimental methods to address different questions in the field of economics. The second chapter is related to the field of cultural economics, it investigates whether rice cultivation that practiced hundreds of years ago give rise to a cooperative social norm that has profound and long lasting effects on individual's behavior in an incentivized and strategic setting. The second chapter investigates whether testosterone leads to reputation enhancing prosocial behavior in the context of charitable donations, which is related to the field of neuroeconomics. In the forth and final chapter, we aim to disentangle the effect of two distinct aspects of reputation concern, namely, pursuit of honor and avoidance of shame, on charitable behavior.

Several governments worldwide have set up measures to promote the electric vehicle (EV) as one option to address environmental issues. However, EV sales figures suggest that the EVs have not been widely accepted by consumers due to cost (notably battery cost), the limited range of products, long recharge times and the limited availability of recharging infrastructure. This indicates the need for in-depth research on the relevant EV supply chains (SCs) to help overcome these barriers. This thesis provides a systematic process for the analysis of EV triple SCs, namely the vehicles, the energy supply and the EV batteries. Three models have been developed for the triple SC-specifically, an EV subsidy model for the vehicle SC; a charging station (CS) selection scheme model for the energy SC; and a closed-loop supply chain (CLSC) model for the EV battery SC. Game theory is used in both EV subsidy model and the EV battery CLSC model, and the stochastic analysis with a Monte-Carlo simulation is used in the CS selection scheme model. Numerical analysis is an inherent element of each of the models to demonstrate their applications. The EV subsidy model involves the government, the gasoline vehicle (GV)/EV manufacturer and retailer and the vehicle customer. The optimum profit for each party and the optimal government subsidy are discussed based on mathematical formulas. The results show that the optimal subsidy is the same, whether it is for the EV manufacturer or the EV customer. In the EV energy supply, the author proposes three EV charging station (CS) selection schemes| per-time selection, bulk selection and combined per-time and bulk selection. Using mathematical analysis and simulation, it can be found that the Per-time selection has the best performance. However, combined selection could be the optimal choice for policymakers as it performs similarly to per-time selection while reducing the system cost. Last, the EV battery CLSC model involves the EV battery manufacturer and remanufacturer and divides the entire life cycle into three periods, the author discusses two optimization decisions: to optimize the profit of each party and to optimize the optimal profit. Although the results and relationships are complex, they can be simplified as being linear or quadratic, which is helpful for the stakeholders in the SC. In summary, this research proposes and discusses the EV triple SC that includes the vehicle, the energy supply and the EV battery. However, the interrelationship between internal sub-supply chains is less studied and will be the direction of future research.

Scholars from Mainland China are increasingly publishing in the medium of English, in order to gain visibility and credibility worldwide. However, the visibility of Chinese scholars in the Social Sciences is strikingly low. Due to the holistic, interpretative, reiterative nature of knowledge in the Social Sciences, writers have to work harder to establish personal credibility through claim-making negotiations, sharing sympathetic understanding and promoting tolerance in their readers (Becher, 1994; Becher & Trowler, 2001; Hyland, 2000). This thesis investigates differences in stance and voice style between scholars from Mainland China and Britain so as to derive new information which might be useful to novice researchers in the Social Sciences (particularly applied linguistics) who intend to publish internationally. A corpus of 30 research articles in applied linguistics was analysed in terms of Appraisal Theory (Martin & White 2005), theory of context (Xu & Nesi, 2017) and genre analysis (Swales 1990, 2004), using the UAM Corpus Tool (O’Donnell 2011). Findings from this analysis suggest that both the Chinese and the British authors are aware of the need to argue for their own opinions and maintain good relationships with their readers, but choose contrasting ways to realize these same purposes. Generally the Chinese authors try to maintain writer-reader relationships by avoiding explicit attitudinal evaluation of the work of others, while the British authors try to maintain writer-reader relationships by toning down or only evoking stance. The Chinese authors argue for their own positions by reinforcing their explicit attitudes, adding multiple references, sharpening the completion of tasks and construing claims as unquestioned, whereas the British authors argue for their own positions by explicitly evaluating people and phenomena. Because the statistically significant differences in stance and voice strategies revealed in this thesis indicate differences between Chinese and British scholars’ argumentative styles, they suggest the need for a new way of perceiving Chinese ethnolinguistic impact on research writing, and might also inform the teaching of academic writing in the social sciences.

This thesis first investigates the moment and memory properties of exponential-type conditional heteroscedasticity models. This primarily includes exponential generalised autoregressive conditional heteroscedastic (EGARCH) models, the fractionally integrated EGARCH model of Bollerslev and Mikkelsen (1996) (FIEGARCH(BM)), the hyperbolic EGARCH (HYEGARCH) model and the FIEGARCH(DL) model, as presented in Chapter 2. The moment conditions of these models are derived from previous literature, and the memory properties are measured by using the near-epoch dependence (NED) functions of an independent process approach. The existence of moments supports the limited memory properties of these models. This study shows that exponential autoregressive conditional heteroscedastic (EARCH)(∞) processes may exhibit geometric memory, hyperbolic memory or long memory. The EGARCH is a case of a geometric memory process. The FIEGARCH(BM) and HY/FIEGARCH(DL) processes can exhibit hyperbolic memory or long memory, depending on the sign of the memory parameter. The study also derives the functional central limit theorem (FCLT) or fractional FCLT for the relevant processes in these exponential-type conditional heteroscedasticity models. Finally, the results of the simulation show that the HYEGARCH model has a hyperbolic memory and that the FIEGARCH(DL) model can capture long memory in absolute return series. Next, the study investigates the asymptotic properties of the quasi-maximum likelihood estimator (QMLE) in autoregressive moving average (ARMA) models with EGARCH or HY/FIEGARCH(DL) errors in Chapter 3. This part of the study aims to investigate the asymptotic theory of the ARMA(1,1)-EGARCH(1,1) models and that of the pure HY/FIEGARCH(DL) models. First, the literature on the asymptotic properties of the ARMA-GARCH and EGARCH processes is reviewed. The conditions for the consistency and asymptotic normality of the QMLE of the ARMA-EGARCH models are then demonstrated. This analysis also provides an investigation of that of the QMLE in the HY/FIEGARCH(DL) processes. A Monte Carlo simulation is used to study the properties of the QMLE in the pure HY/FIEGARCH(DL) processes. Lastly, in a study co-authored with Professor James Davidson, we derive a simple sufficient condition for strict stationarity in the ARCH(∞) class of processes with conditional heteroscedasticity. The concept of persistence in these processes is explored, and is the subject of a set of simulations showing how persistence depends on both the pattern of the lag coefficients of the ARCH model and the distribution of the driving shocks. The results are used to argue that an alternative to the usual method of ARCH/GARCH volatility forecasting should be considered.

China is a huge country with a big territory and an even bigger population. The issue of how to deal with local government is one that has long vexed the central government of every Chinese epoch. Since the establishment of the People’s Republic of China, China has experienced rounds of decentralization and centralization. The central-local relationship has experienced many changes as well. Thus, there is fierce debate over whether China is still a unitary country or whether it has transformed into a federalist nation. The central aim of this thesis is to provide an examination of the nature of the central-local relationship in Chinese governance, and to provide an understanding of how this central-local relationship affects healthcare social policies in China. Research was conducted cumulatively in three steps. First, it drew a lesson from the variety of literature on arrangement of central-local relations and policy process to build a theoretical context and foundation for this research. Secondly, the research examined whether the central-local relationship in China could be categorized as a federalist system or a unitary system, or whether neither label was suitable. This examination led to an understanding of “de facto federalism”, which provides an institutional description of the central-local relationship in Chinese governance. The examination was also conducted with reference to Paul Pierson’s three characteristics of federalism, namely: reservation of specific powers to constituent units; expression of interests to the centre; and the extent of commitment to fiscal equalization across the states. Thirdly, this research conducted a fieldwork study which included a document study and semi-structured interviews in three different Chinese provinces to examine and explore the nature of central-local relation in China and its impact on healthcare policy. It was observed that the de facto federalism partly captures the nature of central-local relationship from perspectives such as the extent of local autonomy and the reciprocity mechanism within central-local interactions. However, the central-local relationship in China does have many unique features. It is profoundly influenced by the culture of unity, which seems to be the boundary of any reform in China. Meanwhile, China has a long history of non-institutionalized central-local power distribution, which is the result of a “dynamic” central-local relationship. The central government takes more initiatives to change the power distribution within such a dynamic central-local relationship. Indeed, de facto federalism has been constrained by these features. The rise of the “project mechanism” could be regarded as a sign of this dynamic relation and the central government seems to take more and more initiative when dealing with local governments.

This thesis aims at providing rigorous micro-foundations for explaining detailed firm and product level trade flows. In this research I investigate the mechanisms by which firms break into foreign markets and the impact of “\(entry\)” on \(growth\), \(further\) \(entry\) and \(exit\), as well as the potential biases when measuring the export growth of firms using annual data. Specifically, I introduce Quarter-Year, Month-Year and Week-Year estimations and examine whether there are certain countries that act as launch pads for Argentinean firms to target exports. My results provide an accurate measure of export growth and have implications for trade policy. I introduce the concept of “trade liberalisation externalities” whereby bilateral trade liberalisation can lead to an expansion of trade to third countries via the process of sequential exporting. This research also deals with the relationship between exporting, importing and product quality.

The employment problem of college graduates has evoked concern from students to politicians. Since 1999, the Chinese government has promoted an enrollment expansion policy evident in the subsequent data which identified an increased college graduation rate. Furthermore, the English language proficiency of Chinese college students was emphasized within their university academic requirements as well as from prospective employers. Consequently, there is increased pressure on Chinese college students to obtain employment after graduation. The objective of this quantitative research study was to investigate the ways in which English language proficiency affects employment attained by Chinese graduates, specifically those in electrical engineering programs. The study framework was based on human capital theory. Within the quantitative data analysis, a structural equation modeling analysis approach was used to validate the research model and test the data. The convenience sample for this study consisted of 128 electrical engineering graduates from one Chinese university. Using maximum likelihood estimation analysis, the final structural model confirmed a good fit for the model. The key findings included the following: the graduates' levels of English language proficiency have a moderate correlation with human capital, human capital has a positively significant impact on employment of graduates, and the graduates' levels of English language proficiency have no significant impact on their employment. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]

The discoveries of materials property-property correlations usually require prior knowledge or serendipity, the process of which can be time-consuming, costly, and labour-intensive. On the other hand, artificial neural networks (ANNs) are intelligent and scalable modelling techniques that have been used extensively to predict properties from materials’ composition or processing parameters, but are seldom used in exploring materials property-property correlations. The work presented in this thesis has employed ANNs combinatorial searches to explore the correlations of different materials properties, through which, ‘known’ correlations are verified, and ‘unknown’ correlations are revealed. An evaluation criterion is proposed and demonstrated to be useful in identifying nontrivial correlations. The work has also extended the application of ANNs in the fields of data corrections, property predictions and identifications of variables’ contributions. A systematic ANN protocol has been developed and tested against the known correlating equations of elastic properties and the experimental data, and is found to be reliable and effective to correct suspect data in a complicated situation where no prior knowledge exists. Moreover, the hardness increments of pure metals due to HPT are accurately predicted from shear modulus, melting temperature and Burgers vector. The first two variables are identified to have the largest impacts on hardening. Finally, a combined ANN-SR (symbolic regression) method is proposed to yield parsimonious correlating equations by ruling out redundant variables through the partial derivatives method and the connection weight approach, which are based on the analysis of the ANNs weight vectors. By applying this method, two simple equations that are at least as accurate as other models in providing a rapid estimation of the enthalpies of vaporization for compounds are obtained.

This thesis focuses on the development of the model-based fault detection and isolation /fault detection and diagnosis (FDI/FDD) techniques using the unknown input observer (UIO) methodology. Using the UI de-coupling philosophy to tackle the robustness issue, a set of novel fault estimation (FE)-oriented UIO approaches are developed based on the classical residual generation-oriented UIO approach considering the time derivative characteristics of various faults. The main developments proposed are: - Implement the residual-based UIO design on a high fidelity commercial aircraft benchmark model to detect and isolate the elevator sensor runaway fault. The FDI design performance is validated using a functional engineering simulation (FES) system environment provided through the activity of an EU FP7 project Advanced Fault Diagnosis for Safer Flight Guidance and Control (ADDSAFE). - Propose a linear time-invariant (LTI) model-based robust fast adaptive fault estimator (RFAFE) with UI de-coupling to estimate the aircraft elevator oscillatory faults considered as actuator faults. - Propose a UI-proportional integral observer (UI-PIO) to estimate actuator multiplicative faults based on an LTI model with UI de-coupling and with added H∞ optimisation to reduce the effects of the sensor noise. This is applied to an example on a hydraulic leakage fault (multiplicative fault) in a wind turbine pitch actuator system, assuming that thefirst derivative of the fault is zero. - Develop an UI–proportional multiple integral observer (UI-PMIO) to estimate the system states and faults simultaneously with the UI acting on the system states. The UI-PMIO leads to a relaxed condition of requiring that the first time derivative of the fault is zero instead of requiring that the finite time fault derivative is zero or bounded. - Propose a novel actuator fault and state estimation methodology, the UI–proportional multiple integral and derivative observer (UI-PMIDO), inspired by both of the RFAFE and UI-PMIO designs. This leads to an observer with the comprehensive feature of estimating faults with bounded finite time derivatives and ensuring fast FE tracking response. - Extend the UI-PMIDO theory based on LTI modelling to a linear parameter varying (LPV) model approach for FE design. A nonlinear two-link manipulator example is used to illustrate the power of this method.

The removal of contamination particles from silicon wafers is critical in the semiconductor industry. Traditional cleaning techniques encounter difficulties in cleaning micro and nanometer-sized particles. A promising method that uses acoustically-driven micro-bubbles to clean contaminated surfaces has been reported. However, little is understood about the microscopic interaction between the micro-bubble and particle. This thesis explores the mechanism underlying the ultrasonic cleaning using micro-bubbles at the micrometer scale. The investigation was carried out from the perspective of bubble dynamics near a surface and bubble-particle interaction. Prior to contributing to the particle removal, micro-bubbles normally need to be transported to a target surface. The motion of a bubble was analyzed based on a force balance model for single and multi-bubble translations respectively. A good agreement is found between the observed bubble movement trajectories and the theoretical predictions. After arriving on a surface, a micro-bubble starts to disturb the flow field near the boundary through its oscillation. The characteristics of the flow field are closely related to the bubble oscillation modes. The influence of a wall on the change of bubble oscillation mode during its translation toward the boundary was studied. The relationship between bubble oscillation modes and the corresponding microstreaming around the bubble was established. The experimental results of bubble oscillation modes and the flow motion are quantitatively in good agreement with the simulation results. From a mechanic point of view, a possible ultrasonic cleaning mechanism is explained by exploring the relationship between different torques that are exerted on micro and sub-micrometer-sized particles. This estimation provides a qualitative insight into the ultrasonic cleaning process at a moderate pressure amplitude. The experimental investigation of the complicated particle detachment process requires improved test equipment to be developed in the future.

This thesis studies the provision of public pension system through three different approaches. Part one focuses on demographic change and pension system reforms in China. It reviews the historical reforms and the problems and suggestions associated with the current system. More importantly, by applying a calibrated overlapping generations general equilibrium simulation model, it investigates the impact of the demographic changes and the choice of pension system to the individual choices and macroeconomic variables in the future. As with all social insurance programs, the provision a public pension system involves a trade-off between protection and distortion. The second part is a theoretical study about the optimal level of public pension system. It derives the optimal pension benefit level by considering the welfare loss imposed by the saving and labour supply distortion. The third part of the thesis, is an empirical study investigating the reasons for different choices in pension systems. There are three types of public pension systems popular throughout the world: Pay-As-You-Go (PAYG), Funded and Mixed. The latter two have grown up largely since 1980s, after Chile successfully built its Funded system. By applying logistic regression tests, we examine the likely social and economic variables which have been affecting the choices.

A critical evaluation of the national profile of energy supply and demand and the associated greenhouse gas (OHO) emissions in China has been conducted. The contribution of the transport sector in China, the road transport sector in particular, to China's overall energy demand and OHO emissions has been assessed and compared with values for other countries. Approaches for reducing energy demand and OHO emissions in the road transport sector worldwide have been reviewed. A detailed bottom-up model has been developed using 'LEAP' software, to estimate future energy demand and OHO emissions in China's road transport sector, incorporating China's recent efforts in alternative fuel promotion. Modelling approach and historical data used have been tested and verified to ensure reliability. Two scenarios have been designed to describe the future strategies relating to the development of China's road transport sector between 2005 and 2030. The 'Business as Usual' scenario is used as a baseline reference scenario, in which the government is assumed to do nothing to influence the long-term trends of road transport energy demand. The 'Best Case' scenario is considered to be the most optimized case where a series of available reduction measures are assumed to be implemented. Energy demand and OHO emissions in China's road transport sector up to 2030 are estimated in these two scenarios. The reduction potential and the relative contribution of each measure have been estimated. A 'life cycle assessment' model for the road transport sector has been developed. The life cycle energy demand and OHO emissions in China's road transport sector are estimated using the model. The reduction potential and the relative contribution of each measure have been re-assessed from a life cycle perspective. Potential impacts on global oil resources, availability and prices are discussed. The importance of life cycle assessment in evaluating the effects of different reduction measures is discussed. Policy implications are presented.

Research has found that effective FPD activities improve faculty's instructional practices and pedagogy, technology skills, and knowledge and that the impact last over time (Rutz, Condon, Iverson, Manduca, & Willett, 2012). FPD activities also reduce job burnout and increase a sense of belonging and morale among faculty (Thomas, 2012). However, many colleges and universities are observing low participation rates, and there is a concern about a mismatch of FPD activities and changing faculty needs. Using a mixed methods research design and an online survey instrument that consisted of Liked scale and open-ended items, this study identified factors that motivate faculty to participate in FPD activities and evaluated faculty perceptions about the value and usefulness of FPD activities. A total of 892 full-time and part-time faculty at nine California State University campuses participated in this study. There were three categories of major findings: (a) relationships between faculty demographics, motivation, and perceived value and usefulness of FPD activities, (b) faculty motivation and perceptions about FPD, and (c) There are positive interrelationships between faculty motivation, perceived value and usefulness of FPD activities. Highlights among the findings included a positive association between off campus FPD involvement with faculty motivation and positive associations between online and off campus FPD with both the perceived value and the perceived usefulness of FPD. The discussion of implications and recommendations included (a) the need for campuses to conduct faculty needs assessments for clear and aligned goals; (b) strategies for planning FPD activities; (c) the use of data to improve FPD effectiveness; (d) the need for policies, resources, and recognitions that promote a culture of FPD; (e) the importance of collaboration; and (f) recommendations for further study related to the regression findings. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]

It has always been an important development strategy for software to keep pace with its evolution and meet stakeholder needs. Therefore, many software systems have very short release cycles. When many releases appear, how to test them becomes more and more important. Since the new features are often innovative, there are no conventional tools to test them, and manual testing is actually very cumbersome, not to mention that there are many non-testable features. This thesis mainly investigates the testability of the 109 features of 16 versions of the Mozilla Firefox browser from May, 2022 to July, 2023. The objects are the features of Firefox browser itself under the Windows 11 operating system, excluding additional plug-ins or Firefox browsers under other operating systems. Through manual testing and analyses, the features of the Firefox browser are divided into testable and non-testable, and the reasons why features are non-testable are given. Since the object being tested is the features of the browser, Selenium—a tool dedicated to testing browsers is considered. In this thesis, the Selenium package based on Python is used to simulate manual testing. By comparing the results of manual testing and the results generated by Selenium, we can gain insights into the limitations of Selenium testing, which contributes to improving automated testing in the future work. In this thesis, we also employ OpenAI’s ChatGPT (version 3.5) to assess the testability of each feature. We then compare ChatGPT’s answers with the manual analysis results.

I test the impact of information acquisition (proxied by SEC EDGAR web traffic) on market informativeness about deal value creation in mergers and acquisitions (M&As). Information acquisition about merging firms, industry rivals, and supply-chain firms enhance short-term market reactions informativeness about long-term merger operating synergies. The effects are more pronounced among M&As with more sophisticated investors, new information, and pre-merger information asymmetry. Furthermore, information acquisition about merging firms improves the market’s assessment of financial synergies. In addition, non-deal related firms with more downloads experience an increase in subsequent takeover probability. The evidence from difference-in-differences analysis and instrumental approach analysis suggests potential casual effects of information acquisition on market reaction informativeness. Overall, this paper demonstrates that information acquisition improves the market’s assessment of value creation in M&As.

Previous studies have contributed to understanding the cognitive process underlying the performance of simultaneous interpreting (SI) in conferences, yet less attention has been paid to courtroom settings, where SI has been extensively employed. SI is crucial in both the International Criminal Court and Australian domestic courts in ensuring the defendant can follow proceedings being conducted in a language they do not speak (Stern, 2011). The importance of SI performance has been highlighted in studies on court interpreting, with several factors being identified as potentially affecting this performance (e.g., Ng, 2023; Stern, 2018). However, few studies have experimented with SI performance in court contexts. The current thesis aims to examine the impact of court-specific factors on SI performance and explain these factors’ effects by mobilising Cognitive Load Theory. This study examines how independent variables (interpreting expertise and prolonged interpreting turns) and their interactions with control variables (input speed and complexity of the source text) affect SI performance and cognitive load (CL) in courtroom contexts. The quality of SI performance was assessed using a refined coding-based system, while CL was measured using hesitations in output interpretations. A natural experiment was conducted with 15 interpreting students and 15 professional interpreters who remotely performed SI in two settings: (1) Environment A, in which interpreters had no access to the speaker’s script and experienced poor acoustic conditions as they practised remote SI in Australian courtrooms, and (2) Environment B, which simulated two working conditions of the International Criminal Court, in which interpreters performed remote SI during the COVID-19 pandemic, namely the advanced provision of the speaker’s script for preparation and favourable acoustic conditions. Using a mixed modelling approach, the general findings show that CL can predict SI performance scores and that interpreting expertise benefits SI performance and reduces CL. Prolonged interpreting turns hinder performance and increase CL, which is reinforced by the increase in input speed and textual complexity. The results of a descriptive analysis conducted to compare the two experimental settings suggest that the participants who worked in an environment with more optimal working conditions experienced less CL and better overcame the negative effects of prolonged interpreting turns on SI performance. However, the outcomes of the examinations of different levels of each independent variable appear to be more complex. This research has provided empirical evidence that enhances the understanding of the impact of task-, environment-, and interpreter-related factors on SI performance and CL in the court context. Additionally, it offers valuable insights into interpreter training and professional practices that align with both national and international standards aimed at improving interpreters’ working conditions.

Graphene, a nanosheet derived from graphite, has been regarded as a highly promising material for supercapacitors due to impressive theoretical surface area of 2600 m2 g−1, chemical stability and exceptional electrical conductivity. As a crucial member of the graphene-based materials family, graphene oxide (GO) has garnered significant attention due to its exceptional ability for chemical functionalization and outstanding stability. The pure GO has a low energy density of activate cites leading to poor powder density. Thus, activating materials are necessary to be added to improve the performance of GO-based supercapacitors. Transition metal oxides (TMOs) are extensively employed as active materials in GO supercapacitors due to their high reactivity. The main aim of this work is to study the variation trend about the influences of synthesis temperature on specific capacitance via combining experimental dataset with machine learning. The thesis is divided into two main sections. In the first section, a comparison was made among the electrochemical performances of four different TMOs doped GO electrode materials. It was observed that cobalt oxide samples exhibited the highest capacitance, prompting further research to characterize their distinct physical properties (such as morphology, composition, and valence) at different temperatures. It was observed that Co(Ⅲ)/Co(Ⅱ) ratio is a key parameter influencing the capacitance. The performance is least when the Co(Ⅲ)/Co(Ⅱ) ratio is closer to the critical value of 0.7. In the second part, we combine the machine learning (ML)-based random forest (RF) model with experimentally observed X-ray photoelectron spectroscopy (XPS) data to construct the complete chemical analysis dataset of Co(Ⅲ)/Co(Ⅱ) ratio for thermally synthesized Co-rGO supercapacitor electrodes. The ML predicted dataset could be further coupled with other experiment results, such as cyclic voltammetry (CV), to establish a precise model for predicting capacitance, with ML coefficient of determination (R2) value of 0.9655. Furthermore, the error between predicted capacitance and experimental validation is found to be less than 8%. Our work indicates that RF can be used to reduce experimental resource consumption for materials analysis. Moreover, the RF-predicted result can be further utilized in experimental and computational analysis.

In the past decade, the proliferation of data and the emergence of large language models have presented both opportunities and challenges in academia. The expanding volume of data, which records knowledge from various human activities, enables data-driven approaches to optimizing numerous aspects of industrial manufacturing and people's daily life. These improvements largely stem from machine learning models trained with this data. However, the industry still faces limitations in both extracting knowledge from large, unstructured, or heterogeneous datasets and transforming the extracted knowledge into actionable insights. This challenge is exacerbated in highly specialized domains where only a few analysts possess the expertise to interpret the data. Despite the recent advancements of large language models providing more intelligent assistance for many data analysis tasks, it remains essential to ensure that these machine learning models and the knowledge they encompass are safe to use and employed for social good with human verification.In my dissertation work, I develop visual analytics (VA) and human-computer interaction (HCI) methodologies for representing and interacting with various forms of knowledge and data, particularly text data. I propose a visual knowledge discovery framework that integrates human expertise with computational approaches throughout the knowledge discovery process, while also addressing the limited availability of domain experts and the increasing scale of data. Moreover, I investigate how visual analytics can efficiently and safely harness extensive knowledge from large machine learning models, enabling users to effectively steer the exploration process and make well-informed decisions. This dissertation presents six published research works organized around my visual knowledge discovery framework and its three key tasks: knowledge exploration, knowledge presentation, and knowledge exploitation. Firstly, I demonstrate how visual analytics can support knowledge exploration with large, high-dimensional, and heterogeneous data in the domain of manufacturing and machine maintenance. Subsequently, I introduce two knowledge presentation solutions for two distinct types of data—numerical data facts and unstructured text data. Lastly, I showcase three visually-assisted knowledge exploitation applications in various domains and scenarios, encompassing document summarization, technical text annotation, and data-driven machine learning model validation.My work demonstrates how mixed-initiative methods through visual analytics applications can resolve real-world challenges in highly-specialized domains. I leverage state-of-the-art machine learning techniques, particularly natural language processing models, while always involving domain practitioners in the loop. My approach facilitates communication among parties with mismatched knowledge levels, including domain experts, data analysts, computer scientists, and artificial intelligence. Meanwhile, I prioritize the critical role of human knowledge and integrate it into intelligent visualization interfaces that undergo qualitative evaluations. I believe that domain experts' insights, supervision, and verification are invaluable, regardless of how advanced machine learning techniques become. Through the projects outlined in this dissertation, I hope to encourage philosophical and social discussions surrounding the rapidly expanding field of artificial intelligence. Ultimately, my objective is to contribute to a future where intelligent visual analytics systems can augment and enhance human capabilities, enabling individuals to navigate through the potential challenges brought by advanced AI techniques.

Porcine epidemic diarrhea virus (PEDV) is an alphacoronavirus. It was initially reported in 1970s in Europe. However, massive porcine epidemic diarrhea (PED) epidemics did not occur in China until late 2010. PEDV was first introduced into the United States (US) and rapidly spread nationwide in 2013. PEDV infection of suckling piglets causes up to 100% mortality and the US pig industry lost 10% of domestic pig population and $0.9 to $1.8 billion during 2013-2014, posing a great threat to the pork industry sustainability. Live attenuated vaccines (LAVs) given orally to pregnant sows/gilts replicate in the intestines and induce lactogenic immunity in sows/gilts. They are the most effective approach to provide passive immunity to neonatal pigs against PED via colostrum and milk. However, safety concerns associated with potential virulence reversion hinder broad application of LAVs. The goal of this dissertation is to identify promising targets in two viral genes and to develop a recombination-resistant platform for the development of effective and safe PEDV LAVs.My first objective was to evaluate whether the exonuclease (ExoN) domain within PEDV nsp14 is a good target for LAV development. Based on the infectious cDNA clone of a highly virulent PEDV strain (icPC22A), eight mutants targeting nsp14 ExoN catalytic sites, zinc finger, or Mg2+-binding site were designed. Only one E191A mutant, carrying the mutation in Mg2+-binding site, was rescued that was characterized by poor growth in Vero or IPEC-DQ cells of the early passages no.1-3 (P1-3) with peak titers of 1.80 ± 0.12 and 1.42 ± 0.16 log10 TCID50/mL, respectively. However, the P4 of E191A was characterized by dramatically improved growth characteristics reaching a high infectious titer (5.55 ± 0.35 log10 TCID50/mL) in Vero cells, like icPC22A. Sequence analysis demonstrated that the introduced mutation site has reverted to wildtype in the P4 virus. To evaluate the pathogenesis of the E191A-P1, 4-5-day-old gnotobiotic pigs were orally inoculated with 100 TCID50/pig of the E191A-P1, icPC22A, or mock. All pigs inoculated with icPC22A developed severe diarrhea and died within 6 days post-inoculation (dpi). In comparison, only 2 pigs (2/3, 66.7%) in the E191A-P1 group showed mild diarrhea at 6 dpi. At 22 dpi, surviving pigs were challenged orally with 106 PFU icPC22A. E191A-P1-inoculated pigs were partially protected from severe diarrhea and high level of viral RNA shedding in feces. Sanger sequencing showed that the viral genomes in the fecal samples from some pigs (1/3, 33.3%) in the E191A-P1 group have lost the E191A mutation, suggesting its instability. Collectively, based on the existing literature data and our own experiments, mutations at the essential functional sites within nsp14-ExoN domain of PEDV were either lethal or genetically unstable. So, the nsp14-ExoN domain is not a good attenuating target for PEDV LAV development.The second objective was to elucidate the role of PEDV nsp1, a potential host interferon (IFN) antagonist, in the development of PEDV LAV candidates. One PEDV nsp1 mutant icPC22A-nsp1-N93/95A (N93/95A) was designed and rescued by targeting the highly conserved residues 93N and 95N of nsp1 of alphacoronaviruses and betacoronaviruses using the reverse genetics system of PEDV strain PC22A. The N93/95A mutant replicated less efficiently (with a peak titer of 5.72 ± 0.83 log10 TCID50/mL in Vero cells and 6.14 ± 0.23 log10 TCID50/mL in LLC-PK cells) than the virulent icPC22A (6.80 ± 0.00 log10 TCID50/mL in Vero cells and 6.63 ± 0.00 log10 TCID50/mL in LLC-PK cells), and it was more sensitive to type I and type III IFNs pre-treatment than icPC22A. Upon viral infection, IPEC-DQ cells infected with N93/95A mutant exhibited increased IFN-β, IFN-λ1, IFN-λ3, and IFN-λ4 mRNA levels than the cells infected with icPC22A. To evaluate the pathogenesis of N93/95A mutant, 5-day-old gnotobiotic piglets were orally inoculated with 100 TCID50/pig of N93/95A mutant, icPC22A, or mock respectively. Pigs in the N93/95A inoculated group showed delayed onset of diarrhea and reduced fecal viral shedding titers (4.80 ± 0.76 log10 TCID50/mL at 2.00 ± 0.00 dpi) compared with the icPC22A-inoculated pigs (5.80 ± 0.33 log10 TCID50/mL at 1.00 ± 0.00 dpi). In addition to the milder clinical signs, decreased mortality rate (25%) was also observed in the N93/95A mutant group than the icPC22A group (100%). At 22 dpi, the surviving pigs were challenged orally with 106 PFU of icPC22A per pig. Pigs in the N93/95A group had no severe diarrhea (0.00 ± 0.00 days) and shorter duration of mild diarrhea (3.33 ± 1.15 days) than the pigs in the mock-challenge group (duration of diarrhea: 7.17 ± 0.98 days and severe diarrhea: 4.50 ± 2.43 days). The N93/95A mutant inoculation also protected pigs from high fecal viral RNA shedding after challenging: pigs in the N93/95A group showed lower peak viral RNA shedding titers (5.57 ± 0.89 log10 copies/mL) than pigs in the mock-challenge group (10.29 ± 0.76 log10 copies/mL). No pigs died in N93/95A group post challenge. However, 33% (2/6) of pigs in the mock-challenge group died. In summary, the N93/95A mutant was attenuated in vivo and induced robust type I and type III IFN responses and protective immunity to virulent PEDV challenging. So, these nsp1 mutation sites N93A and N95A represent feasible targets for the PEDV LAV development.My third objective was to recode the transcriptional regulatory sequences (TRSs) of PEDV to engineer a recombination-resistant platform for LAV development to increase the safety of PEDV LAVs. First, we generated an icPC22A-derived reporter PEDV dORF3-EGFP, in which the ORF3 of icPC22A was replaced by EGFP. Then, using the dORF3-EGFP as a backbone, we designed and rescued one remodeled TRS mutant (RMT) that carries the recoded leader and body TRSs for all ORFs, except the one for EGFP gene. We found that the RMT mutant formed smaller plaques in Vero cells than the dORF3-EGFP. The sgRNAs for structural proteins spike (S), envelop (E), membrane (M) and nucleocapsid (N) were detected from the RMT- and dORF3-EGFP-infected cells. However, the sgRNA for EGFP was detected exclusively from the dORF3-EGFP but not the RMT infections. Although both viruses showed robust N gene expression by immunofluorescent staining, only the dORF3-EGFP- but not RMT-infected cells exhibited strong green fluorescence signal, indicating the efficient expression of EGFP. These results suggest the incompatibility between the wildtype TRS leading EGFP gene and the recoded leader TRS in the 5’-UTR of the RMT genome, resulting in little EGFP expression in the RMT-infected cells. Multi-step growth kinetics results revealed comparable replication efficiency between the RMT mutant and the dORF3-EGFP. The two viruses reached similar peak infectious titers at similar time points. Experimental infection of neonatal gnotobiotic pigs showed that the RMT mutant had similar attenuation phenotype and replication efficiency to the dORF3-EGFP. Challenge of the pigs with the highly virulent parental strain icPC22A at 19 days post-inoculation showed that both RMT and dORF3-EGFP protected the pigs from severe diarrhea and death, indicating that they retained good immunogenicity. To test whether the RMT decreases recombination rates, we co-infected Vero cells or pigs with RMT (or dORF3-EGFP as a positive control) and a PEDV S-INDEL strain Iowa106, whose S1 sequence can be distinguished from that of PC22A strain due to unique insertions and deletions. We only found potential recombinant viruses from the pigs co-infected with dORF3-EGFP and Iowa106, but not the pigs co-infected with RMT and Iowa106. No recombinant viruses were found in the co-infected Vero cells. These results suggested that the RMT mutant may be more resistant to recombination due to the incompatibility between the recoded and wildtype TRSs and can serve as a platform for LAV development by adding additional desired mutations along its genome.In summary, we identified promising targets, N93 and N95 within nsp1, for LAV development. The introduced mutations within nsp1 decreased its IFN antagonism and made the N93/95A mutant more sensitivity to IFN responses. Meanwhile, the mutant was attenuated in neonatal pigs and retained good immunogenicity protecting piglets from severe diarrhea and death. However, the ExoN domain within PEDV nsp14 is critical for its life cycle. Engineered viruses carrying mutations within functional sites of ExoN were highly genetically unstable. Although the crippled nsp14-E191A virus was attenuated and provided partial protection against virulent PEDV challenge, it mutated back both in vitro and vivo. Furthermore, we generated one recombinant PEDV genome by recoding the TRS region which is resistant to recombination event mediated by TRS-driven template switching. The remodeled mutant exhibited similar replication as its backbone both in vitro and in vivo. In a recombination assay by experimental co-infection, incompatibility of recoded and wild-type TRSs of the remodeled mutants may contribute to resistance to recombination. The knowledge obtained from this dissertation may aid in the development of efficient and safe LAVs against PEDV infection.

Parameter identification and estimation from experimental data is an important problem in mathematical model development. Motivated from disease and immunological study, where repeated collection of data from a single subject is impossible and data could be limited, we focus on the parameter identification/estimation of ODE dynamical systems from a single trajectory, on which there are small amount of exactly measured discrete data, particularly considered uniformly spaced in time. This is an inverse problem where we treat the solution map of initial value problem of the ODE system given the parameter as the corresponding forward problem. We put concentration on mathematical establishment of theories on the existence, uniqueness, robustness, uncertainty quantification about the inverse problem, and provide theoretical proofs on some results. Previous study has been completed for such inverse problems of linear systems. We generalize the results to affine dynamical systems, find sufficient and necessary conditions of existence and uniqueness from either full trajectory or discrete data, and then give bounds on maximal allowable uncertainty. There are also results on affine systems that require new proof techniques or never appear in linear systems. We apply the analysis of affine systems into parameter estimation of linear-in-parameters(LIP) systems, which can yield more accurate estimates than those based on finite differences. We then focus on a two-dimensional Lotka-Volterra system as the first example of LIP systems. We sketch diagrams that partially show the continuous dependence of qualitative behavior of inverse problem solution on data. Novel situations also appear for LV systems where data points are compatible with distinct parameter sets corresponding to same or different qualitative types. Finally, we design a numerical algorithm that estimate parameters faster and with less computation time than direct methods, apply to stiff and higher-dimensional linear-in-parameters systems as well as certain nonlinear systems.