Search

Automatic syntactic analysis is essential for extracting useful information from large-scale learner data for linguistic research and natural language processing (NLP). Currently, researchers use standard POS taggers and parsers developed on native language to analyze learner language. Investigation of how such systems perform on learner data is needed to develop strategies for minimizing the cross-domain effects. Furthermore, POS taggers and parsers are developed for generic NLP purposes and may not be useful for identifying specific syntactic constructs such as subcategorization frames (SCFs). SCFs have attracted much research attention as they provide unique insight into the interplay between lexical and structural information. An automatic SCF identification system adapted for learner language is needed to facilitate research on L2 SCFs. In this thesis, we first provide a comprehensive evaluation of standard POS taggers and parsers on learner and native English. We show that the common practice of constructing a gold standard by manually correcting the output of a system can introduce bias to the evaluation, and we suggest a method to control for the bias. We also quantitatively evaluate the impact of fine-grained learner errors on POS tagging and parsing, identifying the most influential learner errors. Furthermore, we show that the performance of probabilistic POS taggers and parsers on native English can predict their performance on learner English. Secondly, we develop an SCF identification system for learner English. We train a machine learning model on both native and learner English data. The system can label individual verb occurrences in learner data for a set of 49 distinct SCFs. Our evaluation shows that the system reaches an accuracy of 84\% F1 score. We then demonstrate that the level of accuracy is adequate for linguistic research. We design the first multidimensional SCF diversity metrics and investigate how SCF diversity changes with L2 proficiency on a large learner corpus. Our results show that as L2 proficiency develops, learners tend to use more diverse SCF types with greater taxonomic distance; more advanced learners also use different SCF types more evenly and locate the verb tokens of the same SCF type further away from each other. Furthermore, we demonstrate that the proposed SCF diversity metrics contribute a unique perspective to the prediction of L2 proficiency beyond existing syntactic complexity metrics.

Colicin A is a plasmid encoded bacteriocin produced by Escherichia coli (E. coli) regarded as a potential antibiotic. Colicin A is unusual in that it requires acidic lipids for its activity and its pore-forming domain (ColA-P) forms an acidic molten globule state. Previously it was shown that surface aspartyl groups are essential for the stability of colicin A in solution. Learning the role of these surface amino acids in folding and stability of Colicin A pore-forming domain requires knowledge of the folded and unfolded states. To do this the aspartates were replaced by alanine, asparagine, glutamate and glutamine respectively. The folded and unfolded states of the protein were probed experimentally by near and far UV CD spectra and NMR 1H-15N HSQC-maps. The urea unfolding processes were studied and analysed by fluorescence emission spectra. This showed the unexpected result that this single domain protein unfolds in two distinct transitions, the first one being destabilized, and the second one being stabilized, by low pH. The thermal denaturation transitions of the protein were measured by differential scanning calorimetry (DSC) and CD spectroscopy at 209 nm and 295 nm. Unexpectedly the DSC data showed that the measured ΔHtr was always half the ΔHvt’H (van’t Hoff) value and possible reasons for this are discussed. The data show that the aspartate to alanine replacements destabilize the protein significantly and the asparagine mutations do not. ColA-P D216, D274 and D372 are N-cap residues in the full length protein and this may explain why protonated Asp destabilizes the protein at low pH. The glutamate and glutamine mutants stabilize the protein more than expected for true N-capping residues. The biochemical function of the surface aspartates were studied comparing the NMR 1H-15N HSQC-maps of the proteins. The NMR experiments show that the aspartate to alanine mutation destabilize the protein in the same way as the acidic environment.

One of the most controversial issues for long-term stock performance is whether the presence of anomalies is against the efficient market hypothesis. The methodologies to measure abnormal returns applied in the long-run event studies are questioned for their reliability and specification. This thesis compares three major methodologies via a simulation process based on the UK stock market over a period of 1982 to 2008 with investment horizons of one, three and five years. Specifically, the methodologies that are compared are the event-time methods based on models (Chapter 3), the event-time methods based on reference portfolios (Chapter 4), and the calendar-time methods (Chapter 5). Chapter 3 covers the event-time approach based on the following models which are used to estimate normal stock returns: the market-adjusted model, the market model, the capital asset pricing model, the Fama-French three-factor model and the Carhart four-factor model. The measurement of CARs yields misspecification with higher rejection rates of the null hypothesis of zero abnormal returns. Although the application of standard errors estimated from the test period improves the misspecification, CARs still yield misspecified test statistics. When using BHARs, well-specified results are achieved when applying the market-adjusted model, capital asset pricing model and Fama-French three-factor model over all investment horizons. It is important to note that the market model is severely misspecified with the highest rejection rates under both measurements. The empirical results from simulations of event-time methods based on reference portfolios in Chapter 4 indicate that the application of BHARs in conjunction with p-value from pseudoportfolios is appropriate for application in the context of long-run event studies. Furthermore, the control firm approach together with student t-test statistics is proved to yield well-specified test statistics in both random and non-random samples. Firms in reference portfolios and control firms are selected on the basis of size, BTM or both. However, in terms of power of test, these two approaches have the least power whereas the skewness-adjusted test and bootstrapped skewness-adjusted test have the highest power. It is worth noting that when the non-random samples are examined, the benchmark portfolio or control firm needs to share at least one characteristic with the event firm. The calendar-time approach is suggested in the literature to overcome potential issues with event-time approaches like overlapping returns and calendar month clustering. Chapter 5 suggests that both three-factor and four-factor models present significant overrejections of the null hypothesis of zero abnormal returns under an equally-weighted scheme. Even for stocks under a value-weighted scheme, the rejection rate for small firms shows overrejection. This indicates the small size effect is more prevalent in the UK stock market than in the US and the calendar-time approach cannot resolve this issue. Compared with the three-factor model, the four-factor model, despite its higher explanatory power, improves the results under a value-weighted scheme. The ordinary least squares technique in the regression produces the smallest rejection rates compared with weighted least squares, sandwich variance estimators and generalized weighted least squares. The mean monthly calendar time returns, combining the reference portfolios and calendar time, show similar results to the event-time approach based on reference portfolios. The weighting scheme plays an insignificant role in this approach. The empirical results suggest the following methods are appropriately applied to detect the long-term abnormal stock performance. When the event-time approach is applied based on models, although the measurement of BHARs together with the market-adjusted model, capital asset pricing model and Fama-French three-factor model generate well-specified results, the test statistics are not reliable because BHARs show severe positively skewed and leptokurtic distribution. Moreover, the reference portfolios in conjunction with p-value from pseudoportfolios and the control firm approach with student t test in the event-time approach are advocated although with lower power of test. When it comes to the calendar-time approach, the three-factor model under OLS together with sandwich variance estimators using the value-weighted scheme and the mean monthly calendar-time abnormal returns under equal weights are proved to be the most appropriate methods.

Chronic conditions are the leading cause of morbidity and mortality in the world. Traditional health care for people suffering from a chronic condition is expensive and consumes significant resources. This Thesis focuses on investigating novel applications of technologies and services to improve the support offered to patients in the self-management of their chronic conditions. A decision support framework for self-management is proposed. It combines three theoretical perspectives: utilizing information gleaned from behavioural, psychological and biological interventions. A personal self-management system (PSMS), developed within SMART2, utilizes such a framework. It facilitates the integration of activity information, self-reporting, vital signs and lifestyle monitoring data. Approaches have been developed and applied to analyse the information collected and provide feedback to the 'patient' to assist with self-management. An orientation free adaptive step detection (OF ASD) algorithm was developed, deployed and evaluated with a smart phone equipped with accelerometers to detect levels of activity in the form of human steps. The advantages of the OF ASD algorithm following evaluation were found to be the self adaptive nature of the approach for each individual and the orientation free notion for smart phone placement. The accuracy and sensitivity of the OF ASD algorithm achieved performance rates of 93.4% and 93.2%, respectively. Furthermore, an approach was considered involving the identification of a patient's health status through self-reporting. Several feature selection and classification methods were evaluated and the optimal combination for the purposes of classification was reported. The results have indicated that a Multi layer Perceptron (MLP) based classifier had the best classification performance on an optimised subset of 10 questions. Based on the study it has found that the analysis of the lifestyle of a patient produces relevant knowledge that can be used to inform and promote behavior change on the part of the patient and to support self-management. Overall, the results from the research have demonstrated that assistive technologies and suitable algorithms have the capability to facilitate self-management for people with chronic conditions.

Glucagon-like peptide-1 (GLP-1) enhances glucose-dependent insulin secretion and promotes β-cell function via its receptor (GLP-1R), which therefore is a validated target for the treatment of type 2 diabetes. Due to difficulties with peptide therapeutics, it is important to find small-molecule GLP-1R agonists. This leads to a need to understand the structure, function and regulation of the receptor, particularly, differences between agonisms mediated by GLP-1 (orthosteric agonist) and small molecules. The GLP-1R contains a putative N-terminal signal peptide sequence, which is assessed here by recombinantly expressing several epitope-tagged GLP-1R constructs in HEK293 cells. The findings demonstrate that the GLP-1R is expressed predominately at the plasma membrane and also slightly cytosolic. Only fully glycosylated, mature form of the receptor is able to traffic to the cell surface and performs the function. The signal peptide sequence of the GLP-1R is essential for synthesis. After fulfilling the function, this sequence is cleaved and thus not part of the mature protein. The cleavage of signal peptide is critical for processing and trafficking of the GLP-1R. Based on one of these constructs generated here, a cell line (HEK293: GLP-1R-EGFP) with stable expression of the visible GLP-1R is established, which allows observations and determinations for ligand-mediated receptor internalisation in real time. Compound 2 (6,7-dichloro-2-methylsulfonyl-3-N-tert-butylaminoquinoxaline) has been described as a GLP-1R allosteric modulator and agonist. Findings here that compound 2-mediated agonisms on both the wild-type (WT) GLP-1R and the mutant with removal of the N-terminal domain provide direct evidence for the allosteric agonism. Interestingly, compound 2-mediated cAMP response is enhanced by orthosteric antagonist exendin 9-39, but the latter inhibits receptor internalisation mediated by compound 2. Recently, it has been hypothesised that the binding of GLP-1 allows a sequence of NRTFD (Asn63-Asp67) in the N-terminus of the GLP-1R to interact with another part of the receptor and cause agonism. This was examined here by generating receptor mutants and synthetic peptides. Findings here that Asp67 plays a key role in stabilising the N-terminal structure of the GLP-1R and thus is critical for processing and trafficking of the receptor protein do not support such hypothesis although synthetic NRTFD mediates a weak and partial agonism on the WTGLP-1R.

Retinal prostheses require the presence of viable population of cells in the inner retina. Evaluations of retina with Age-Related Macular Degeneration (AMD) and Retinitis Pigmentosa (RP) have shown a large number of cells remain in the inner retina compared with the outer retina. Therefore, vision loss caused by AMD and RP is potentially treatable with retinal prostheses. Photostimulation based retinal prostheses have shown many advantages compared with retinal implants. In contrary to electrode based stimulation, light does not require mechanical contact. Therefore, the system can be completely external and does not have the power and degradation problems of implanted devices. In addition, the stimulating point is flexible and does not require a prior decision on the stimulation location. Furthermore, a beam of light can be projected on tissue with both temporal and spatial precision. This thesis aims at finding a feasible solution to such a system. Firstly, a prototype of an optoelectronic stimulator was proposed and implemented by using the Xilinx Virtex-4 FPGA evaluation board. The platform was used to demonstrate the possibility of photostimulation of the photosensitized neurons. Meanwhile, with the aim of developing a portable retinal prosthesis, a system on chip (SoC) architecture was proposed and a wide tuning range sinusoidal voltage-controlled oscillator (VCO) which is the pivotal component of the system was designed. The VCO is based on a new designed Complementary Metal Oxide Semiconductor (CMOS) Operational Transconductance Amplifier (OTA) which achieves a good linearity over a wide tuning range. Both the OTA and the VCO were fabricated in the AMS0.35 μm CMOS process. Finally a 9X9 CMOS image sensor with spiking pixels was designed. Each pixel acts as an independent oscillator whose frequency is controlled by the incident light intensity. The sensor was fabricated in the AMS 0.35 μm CMOS Opto Process. Experimental validation and measured results are provided.

Territorial disputes in and over the South China Sea (SCS) are often regarded as a dangerous flashpoint in the Indo-Pacific with potentially serious global consequences. In the context of this territorial rivalry, tourism has been deployed by several littoral countries as a soft, civilian tactic to enact maritime territorialisation. Situated against this backdrop, this thesis serves as the first empirically-grounded and theoretically-engaged research to analyse the dynamic nexus between tourism and geopolitical territorialisation in the SCS. The thesis takes China’s tourism in the Xisha region as a case study and applies an ethnographic approach. More specifically, by moving away from a state-centric focus, this thesis integrates multiple scales of analysis to comprehensively examine tourism’s entanglements with the state’s practical and formal geopolitics and explore how tourism territorialisation is played out and experienced in an everyday context. For the territorial role and effect of tourism, this thesis presents a theoretical argument that tourism ought to be treated as a performative force of producing nationalised space and as an assemblage element affecting individuals’ geopolitical subjectivation in daily life. In a practical sense, I suggest that tourism should be considered as a constitutive dimension of the geopolitical restructuring processes in the SCS and as a dynamic geopolitical process on its own, which is informed by, and at the same animates, the multiple scales of geopolitical relations. The research findings presented in this thesis show that tourism is more than an enterprise monopolised and unilaterally pursued by Beijing to achieve the narrow pursuit of state-led territorialisation of the SCS. A range of state, substate, non-state, and even non-human, actors are found to participate in and inform maritime territorialisation through tourism. These diverse actors generally have their own agendas, interests, and pursuits in tourism development, which are not always in line with the state intentions and official prescriptions. Through process-tracking the development history of tourism, this thesis shows that China’s launch of tourism in Xisha is precisely a product resulting from domestic political and economic imperatives and stimulations of the then regional geopolitical tensions. Domestically, the Hainan provincial government is found to be an important sub-state actor who introduced the tourism territorialisation idea into the state’s maritime policy and continues to exert significant influence on the implementation and operation of Beijing’s tourism strategy in the region. Notably, the research demonstrates that China’s SCS tourism is not a static undertaking by identifying a prominent evolution of the roles of tourism in state maritime policy. Since the implementation of the ‘One Belt and One Road Initiative’ in late 2013, Beijing has gradually positioned tourism as a potential peacemaker for facilitating pan-SCS economic cooperation. Investigating how tourism is concretely practised and performed on the ground by state-owned tourism enterprises, another important substate-actor, the research sheds light on how territoriality and sovereignty are enacted, materialised, and performed on a daily basis. The research reveals Xisha tourism has been largely normalised as a self-sustaining corporate operation with government intervention decreasing over time. Despite being subordinated to governments, these state-owned tourism enterprises are far from docile policy-undertakers, but are calculating entities seeking to maximise their corporate interests through the tourism industry. Through a closer study of specific tour structures and patterns of organisation, the discourses, representation, performances, materiality, and spatial-temporal techniques of control employed by tourism operators, the thesis illustrates that Xisha tourism is carefully vetted, politically oriented, highly regimented, deliberately managed, elaborately designed, and performatively constituted in order to better curate tourists’ experiences and, in doing so, better produce territory. Tourism operators politicise the tourism landscape of the SCS, thereby choreographing the entire destination to be a complete geopolitical stage. Furthermore, the thesis interrogates the micro-geopolitical experiences and encounters of tourists, investigating the territorial effects and affects of tourism on the personal level. For this, a specific concept, ‘territorial socialisation’, is developed to capture such an effect. The research demonstrates that tourists are pragmatic geopolitical actors rather than passive pawns of the state. Their geopolitics experiences through tourism are connected to, and embedded in, their extensive daily perceptions and experiences of the benefits of China’s national rising and of the unjust international orders on handling territorial disputes. By translating the remote ‘imagined homeland’ into real existence and enabling an intimate, tangible, and affirmative encounter with national territories, tourism works on the emotional, embodied, and cognitive dimensions of territorial socialisation. Importantly, the vast and flowing physicality of the sea is found to complicate tourists’ maritime territorial identity. In addition to the more apparent political and economic schemes behind the development and maintenance of SCS tourism, and in view of the intimate relations between everyday tourism practices and geopolitics, this thesis suggests that tourism is a mundane and intimate way in which people regularly live geopolitics in their daily lives. Together, the empirical discussion of the thesis demonstrates the agency of tourism in configuring the geopolitical process and shows the mutual constitution of tourism and geopolitics in general and the interdependence of tourism and territorialisation in the SCS in particular. By unravelling the multiple facets of the geopolitics of China’s SCS tourism, this thesis contributes to the scholarship on tourism geopolitics, maritime territorialisation, territorial nationalism, and the SCS dispute. It also offers practical insights into the potential ramifications of tourism development, and its implications for our understanding of the geopolitics of the SCS.

Resource allocation in modern wireless networks is constrained by increasingly stringent real-time requirements. Such real-time requirements typically come from, among others, the short coherence time on a wireless channel, the small time resolution for resource allocation in OFDM-based radio frame structure, or the low-latency requirements from delay-sensitive applications. An optimal resource allocation solution is useful only if it can be determined and applied to the network entities within its expected time. For today's wireless networks such as 5G NR, such expected time (or real-time requirement) can be as low as 1 ms or even 100 μs. Most of the existing resource optimization solutions to wireless networks do not explicitly take real-time requirement as a constraint when developing solutions. In fact, the mainstream of research works relies on the asymptotic complexity analysis for designing solution algorithms. Asymptotic complexity analysis is only concerned with the growth of its computational complexity as the input size increases (as in the big-O notation). It cannot capture the real-time requirement that is measured in wall-clock time. As a result, existing approaches such as exact or approximate optimization techniques from operations research are usually not useful in wireless networks in the field. Similarly, many problem-specific heuristic solutions with polynomial-time asymptotic complexities may suffer from a similar fate, if their complexities are not tested in actual wall-clock time. To address the limitations of existing approaches, this dissertation presents novel real- time solution designs to two types of optimization problems in wireless networks: i) problems that have closed-form mathematical models, and ii) problems that cannot be modeled in closed-form. For the first type of problems, we propose a novel approach that consists of (i) problem decomposition, which breaks an original optimization problem into a large number of small and independent sub-problems, (ii) search intensification, which identifies the most promising problem sub-space and selects a small set of sub-problems to match the available GPU processing cores, and (iii) GPU-based large-scale parallel processing, which solves the selected sub-problems in parallel and finds a near-optimal solution to the original problem. The efficacy of this approach has been illustrated by our solutions to the following two problems. • Real-Time Scheduling to Achieve Fair LTE/Wi-Fi Coexistence: We investigate a resource optimization problem for the fair coexistence between LTE and Wi-Fi in the unlicensed spectrum. The real-time requirement for finding the optimal channel division and LTE resource allocation solution is on 1 ms time scale. This problem involves the optimal division of transmission time for LTE and Wi-Fi across multi- ple unlicensed bands, and the resource allocation among LTE users within the LTE's "ON" periods. We formulate this optimization problem as a mixed-integer linear pro- gram and prove its NP-hardness. Then by exploiting the unique problem structure, we propose a real-time solution design that is based on problem decomposition and GPU-based parallel processing techniques. Results from an implementation on the NVIDIA GPU/CUDA platform demonstrate that the proposed solution can achieve near-optimal objective and meet the 1 ms timing requirement in 4G LTE. • An Ultrafast GPU-based Proportional Fair Scheduler for 5G NR: We study the popular proportional-fair (PF) scheduling problem in a 5G NR environment. The real-time requirement for determining the optimal (with respect to the PF objective) resource allocation and MCS selection solution is 125 μs (under 5G numerology 3). In this problem, we need to allocate frequency-time resource blocks on an operating channel and assign modulation and coding scheme (MCS) for each active user in the cell. We present GPF+ — a GPU based real-time PF scheduler. With GPF+, the original PF optimization problem is decomposed into a large number of small and in- dependent sub-problems. We then employ a cross-entropy based search intensification technique to identify the most promising problem sub-space and select a small set of sub-problems to fit into a GPU. After solving the selected sub-problems in parallel using GPU cores, we find the best sub-problem solution and use it as the final scheduling solution. Evaluation results show that GPF+ is able to provide near-optimal PF performance in a 5G cell while meeting the 125 μs real-time requirement. For the second type of problems, where there is no closed-form mathematical formulation, we propose to employ model-free deep learning (DL) or deep reinforcement learning (DRL) techniques along with judicious consideration of timing requirement throughout the design. Under DL/DRL, we employ deep function approximators (neural networks) to learn the unknown objective function of an optimization problem, approximate an optimal algorithm to find resource allocation solutions, or discover important mapping functions related to the resource optimization. To meet the real-time requirement, we propose to augment DL or DRL methods with optimization techniques at the input or output of the deep function approximators to reduce their complexities and computational time. Under this approach, we study the following two problems: • A DRL-based Approach to Dynamic eMBB/URLLC Multiplexing in 5G NR: We study the problem of dynamic multiplexing of eMBB and URLLC on the same channel through preemptive resource puncturing. The real-time requirement for determining the optimal URLLC puncturing solution is 1 ms (under 5G numerology 0). A major challenge in solving this problem is that it cannot be modeled using closed-form mathematical expressions. To address this issue, we develop a model-free DRL approach which employs a deep neural network to learn an optimal algorithm to allocate the URLLC puncturing over the operating channel, with the objective of minimizing the adverse impact from URLLC traffic on eMBB. Our contributions include a novel learning method that exploits the intrinsic properties of the URLLC puncturing optimization problem to achieve a fast and stable learning convergence, and a mechanism to ensure feasibility of the deep neural network's output puncturing solution. Experimental results demonstrate that our DRL-based solution significantly outperforms state-of-the-art algorithms proposed in the literature and meets the 1 ms real-time requirement for dynamic multiplexing. • A DL-based Link Adaptation for eMBB/URLLC Multiplexing in 5G NR: We investigate MCS selection for eMBB traffic under the impact of URLLC preemptive puncturing. The real-time requirement for determining the optimal MCSs for all eMBB transmissions scheduled in a transmission interval is 125 μs (under 5G numerology 3). The objective is to have eMBB meet a given block-error rate (BLER) target under the adverse impact of URLLC puncturing. Since this problem cannot be mathematically modeled in closed-form, we proposed a DL-based solution design that uses a deep neural network to learn and predict the BLERs of a transmission under each MCS level. Then based on the BLER predictions, an optimal MCS can be found for each transmission that can achieve the BLER target. To meet the 5G real-time requirement, we implement this design through a hybrid CPU and GPU architecture to minimize the execution time. Extensive experimental results show that our design can select optimal MCS under the impact of preemptive puncturing and meet the 125 μs timing requirement.

The 2016 Epilepsy Innovation Institute (Ei2) community survey reported that unpredictability is the most challenging aspect of seizure management. Effective and precise detection, prediction, and localization of epileptic seizures is a fundamental computational challenge. Utilizing epilepsy data from multiple epilepsy monitoring units can enhance the quantity and diversity of datasets, which can lead to more robust epilepsy data analysis tools. The contributions of this dissertation are two-fold. One is the implementation of a temporal query for epilepsy data; the other is the machine learning approach for seizure detection, seizure prediction, and seizure localization. The three key components of our temporal query interface are: 1) A pipeline for automatically extract European Data Format (EDF) information and epilepsy annotation data from cross-site sources; 2) Data quantity monitoring for Epilepsy temporal data; 3) A web-based annotation query interface for preliminary research and building customized epilepsy datasets. The system extracted and stored about 450,000 epilepsy-related events of more than 2,497 subjects from seven institutes up to September 2019. Leveraging the epilepsy temporal events query system, we developed machine learning models for seizure detection, prediction, and localization. Using 135 extracted features from EEG signals, we trained a channel-based eXtreme Gradient Boosting model to detect seizures on 8-second EEG segments. A long-term EEG recording evaluation shows that the model can detect about 90.34% seizures on an existing EEG dataset with 961 hours of data. The model achieved 89.88% accuracy, 92.32% sensitivity, and 84.76% AUC based on the segments evaluation. We also introduced a transfer learning approach consisting of 1) a base deep learning model pre-trained by ImageNet dataset and 2) customized fully connected layers, to train the patient-specific pre-ictal and inter-ictal data from our database. Two convolutional neural network architectures were evaluated using 53 pre-ictal segments and 265 continuous hours of inter-ictal EEG data. The evaluation shows that our model reached 86.79% sensitivity and 3.38% false-positive rate. Another transfer learning model for seizure localization uses a pre-trained ResNext50 structure and was trained with an image augmentation dataset labeling by fingerprint. Our model achieved 88.22% accuracy, 34.99% sensitivity, 1.02% false-positive rate, and 34.3% positive likelihood rate.

In the era of context-aware services, users are enjoying remarkable services based on data collected from a multitude of users. To receive services, they are at risk of leaking private information from adversaries possibly eavesdropping on the data and/or the un--trusted service platform selling off its data. Malicious adversaries may use leaked information to violate users' privacy in unpredictable ways. To protect users' privacy, many algorithms are proposed to protect users' sensitive information by adding noise, thus causing context-aware service quality loss. Game theory has been utilized as a powerful tool to balance the tradeoff between privacy protection level and service quality. However, most of the existing schemes fail to depict the mutual relationship between any two parties involved: user, platform, and adversary. There is also an oversight to formulate the interaction occurring between multiple users, as well as the interaction between any two attributes. To solve these issues, this dissertation firstly proposes a three-party game framework to formulate the mutual interaction between three parties and study the optimal privacy protection level for context-aware services, thus optimize the service quality. Next, this dissertation extends the framework to a multi-user scenario and proposes a two-layer three-party game framework. This makes the proposed framework more realistic by further exploring the interaction, not only between different parties, but also between users. Finally, we focus on analyzing the impact of long-term time-serial data and the active actions of the platform and adversary. To achieve this objective, we design a three-party Stackelberg game model to help the user to decide whether to update information and the granularity of updated information.

The Epilepsy Innovation Institute (Ei2) 2016 community survey reported that unpredictability is the most challenging aspect for seizure management. Effective and precise prediction of epileptic seizure onset is now a fundamental computational challenge, and a quick, reliable, and accurate prediction within a shorter lead time can help patients and doctors to take effective actions before the seizure onset to reduce adverse outcomes.We automatically extract European Data Format (EDF) information and epilepsy annotation data from a cross-site source. A web-based annotation query interface is developed to build customized epilepsy datasets. We also introduce a two-phased approach which consists of a deep learning model with Long short-term memory (LSTM) neurons and a prediction memory queue. The LSTM model is applied to detect pre-ictal and inter-ictal periods from EEG signals, and the prediction memory queue is utilized to forecast seizures. Our predictor can output a warning for a potential seizure within a 5-minute lead time, which is significantly reduced from the 60 minute lead time used for the 2014 seizure prediction competition on kaggle.com. We construct patient-specific datasets and cohort-based datasets leveraging the Center for SUDEP Research's (CSR) large-scale labeled epilepsy electroencephalography (EEG) and electrocardiography (ECG) data. The evaluation shows that our model reached an average sensitivity of 95.24%, mean time in warning of 63.15%, and an improvement of 32.09% over a random predictor, which indicates that our deep learning approach is a promising direction for realistic seizure forecasting.

The ability to analyze gene expression data has had a fundamental impact in the biological sciences and on our understanding of the causes and mechanisms of disease. However, a significant statistical challenge is posed by the combination of the small number of replicates together with the large number of genes leading to an undesirable level of misclassified genes when identifying genes with differential expression levels. When multiple gene expression data sets are generated under the same set of experimental conditions, the ques- tion arises as to how to efficiently combine this information. Several methods in the literature have been suggested to aggregate ranked data from multiple sources. We introduce a new classifier, underpinned by Bayesian principles, called Peer Reinforced Ranker (PR-Ranker) which uses density estimation to approximate the probability that a gene is differentially expressed given a collection of ranked lists.Our classifier is amenable to theoretical analysis when the number of genes and lists is large using the theory of large deviations. Under modest technical assumptions we show that asymptotically PR-Ranker has the smallest loss of any rank aggregation procedure. Moreover, we prove that other more ad hoc methods, such as Borda, have a strictly higher asymptotic rate of loss.While the theoretical results are asymptotic, we perform a series of simulation studies that demonstrate that our classifier outperforms existing methods on datasets of realistic size for biological data. Furthermore, we show that the outperformance is even greater when the lists exhibit varying levels of noise or when some sources are corrupted. PR-Ranker automatically adapts to varying data quality and efficiently combines the data from different sources. Finally we apply PR-Ranker to a gene expression data set in a preeclampsia study. The top ranked genes identified were known to be biologically relevant to preeclampsia and our method achieved a substantially higher Consistency Index than other rank aggregation procedures.

Animals face higher predation risk during mating due to temporary conspicuous appearance, physical restriction and lowering of vigilance. Studies were done on animals to investigate mating-related risks and escape strategies. However, few has dug into the internal state of mating animals, and how mating lowers the awareness of danger. In this study, we used Drosophila melanogaster as a model to study the mechanism behind the effect of mating on escape behaviors. We created an environment that allows flies to exhibit both escape and mating behavior, and established a reference showing escapes of wildtype female flies during mating are suppressed. In an attempt to understand the suppression location in the brain of mating on escape behaviors, we optogenetically activated transgenic flies during mating. Our results suggest that activating lobular columnar cells LC6, or giant fiber neurons (GF) do not affect the suppressed escape behavior during mating, indicating that short takeoff escapes are suppressed at or downstream of GF neurons, and for long takeoff escapes, the suppression site is at or downstream of LC6.

Lymph nodes (LNs) are immunological hubs where antigen and naïve lymphocytes meet to mount antigen specific immune responses. In cancers, the tumour-draining lymph node (TDLN) is the site of initial anti-tumour immune response. The TDLN undergoes dramatic reprogramming in response to tumour drainage that is still partly unclear. In vivo models do not easily permit us to follow events longitudinally or to manipulate the environment, while currently there is a lack of in vitro platforms that recapitulate the LN complexity to study its role in response to antigens and in disease. First, I have tried to adapt an existing PDMS-based device to model the LN. Having encountered some limitations, I have then developed a novel hydrogel-based LN-on-chip device. I have optimized a fluidic system, cell culture conditions and hydrogel composition to obtain long term culture of lymphoid primary cultures from spleen with a stromal cell line of fibroblastic reticular cells (FRCs). Different hydrogel formulations were tested and characterized with multiple techniques focusing on maintaining physiological cell morphology in the new 3D environment. In addition, a LN-like architecture was achieved where B cells are segregated from T cells, into B cell follicles. After optimization of the 3D cellular system with built-in fluidics, I applied the model to different context. Initially, PMA/Ionomycin, a strong cell activator, was perfused to obtain a general activation response. Then, LPS was used to trigger a reaction to a bacteria-derived antigen. Finally, using the OT-I model, I have used the device to trigger an antigen specific immune response to an OVA peptide. In the final part of the thesis, I modelled a TDLN in vitro. The LN device was incubated with tumour conditioned media (TCM) and IL-7 downregulation was observed, similarly to in vivo murine melanoma models. Then, to increase the complexity of the source of the tumour cues, a new tumour model was developed through 3D bioprinting, which also includes a shell of cancer associated fibroblasts (CAFs) and immune cells to form a complex engineered tumour microenvironment (TME). Incubation of the LN device with conditioned medium from the engineered TME also induced downregulation of IL-7. v Unlike current in vivo models that require node dissection at each discrete time point, this system allowed the monitoring of tissue remodelling in real-time via live imaging and measure molecular changes at a surface markers level as well as gene expression. Overall, this platform can represent a valid tool to study in vitro the LN complexity in a more tractable way.

3D printing has emerged as a promising tool for on-demand and rapid fabrication of materials. The field of soft material printing typically utilises inks that exhibit viscoelastic properties with elastic moduli in the kPa - MPa range, such as hydrogels and elastomers. Although soft material printing has been frequently used for creating biomimetic mini tissues, its ability to imitate organ functions as a direct result of organ anatomy is yet to be fully realised, and continued innovation in printing method and flexible machinery are needed to drive the field forward. My PhD thesis focuses on advancing the field of soft material printing. Specifically, there are three main scopes in my work. Firstly, I developed an affordable and fully customisable extrusion-based printing platform for soft materials. The platform is equipped with multiple printheads for heterogeneous construct printing, and heating systems and a UV module for tuning the material rheology during and after printing. A detailed assembly instruction and the software design are provided, hence new users can facilely replicate the platform and contribute to the continued development of the platform. In summary, it is anticipated that this entirely hackable platform can facilitate the widespread adoption of the technology, overcoming the cost and flexibility barriers presented in commercial systems. Secondly, to realise the potential of 3D printing for imitating physiological phenomena related to anatomical structures, I created 3D printed cochleae that exhibit similar electro-anatomical features resembling human cochleae. These biomimetic cochlear models were integrated with machine learning to advance clinical predictions of 'current spread' for cochlear implant (CI) patients. The co-modelling framework demonstrated autonomous predictions of patient electric field imaging profile or cochlear geometry, unfolded the electro-anatomical factors causing CI stimulus spread, assisted on-demand printing for CI testing, and inferred patients' in vivo cochlear tissue resistivity by CI telemetry. This framework might facilitate physical modelling and digital twin innovations for neuromodulation implants in healthcare. Lastly, I demonstrate the high flexibility and versatile functionalities of the custom-made 3D extrusion printing platform. Apart from 3D CAD models, the standard geometry input used in 3D printing, the platform accepts unconventional geometry inputs to suit different needs, including coordinates, equations and pictures. Advanced operations, such as liquid dispensing, printing with variable speed and non-planar printing, are permitted with the platform. With the aid of support baths, heating and UV tools, a wide variety of soft materials, including naturally derived hydrogels, pH-responsive hydrogels and elastomers, were successfully printed using the platform. Overall, the perspective provided in this work might guide new users to efficiently design printing processes for soft materials that do not possess suitable rheological and mechanical properties for creating 3D structures with conventional extrusion methods.

Recent advances in biofabrication have unveiled opportunities to build up biomimetic scaffolds and architectures for applications ranging from tissue engineering to bioinspired soft robotics. The structural and functional analogues of ECMs could be seen to structurally exist with two functional phases, the fibril architecture and the hydrated interstitial gel; and extensive efforts have been devoted in recapitulating these two phases by fibres and hydrogels respectively. With the above overarching motivation, this PhD project focuses on the fabrication of a synthetic medical hydrogel, as well as the investigation of cell dynamics on ECM-mimic fibrils. Firstly, a new hydrogel printing strategy, Hydrogel Embedding with Reversible Multiphase Encapsulating System (HERMES), was developed to 3D print chemically-inert and mechanically-robust polyacrylamide (PAAM) based hydrogels. In particular, pre-crosslinked alginate slurries were developed as the framework ink to bind with PAAM precursors and achieve extrusion printing; and novel oil-phase fugitive hydrophobic supporting inks were developed in order to achieve PAAM embedded-printing. HERMES strategy was employed to create volumetric hydrogel structures with geometric complexity beyond existing limitations in casting, and is applicable in printing various other hydrophilic polymers with low viscosity and slow-gelling properties. Secondly, to mimic the morphological characteristics of fibrillar matrices, low-voltage continuous electrospinning techniques was adapted to construct straight, wavy, looped and grid fibre patterns made of polystyrene. With microfibres deposited onto non-passivated surfaces, cells were able to dynamically adapt their shapes in response to the directly-adhered fibre, as well as to the neighbouring patterns; and such adaptation reflect their ability to shuttle and alter between fibre tracks. Key morphological features such as the variation of cells' minor axis were identified for understanding cell migration in fibril matrices, as well as designing future ECM-mimic fibril models for cell migration studies. Taken together, the two parts of work have shown the feasibility of building cell-guiding hybrid architectures made of fibre-integrated hydrogels. This was the ultimate goal but due to Covid-19 pandemic restrictions, the two strategies were demonstrated separately without integration. It is envisaged that this PhD work might lay down the foundation to apply additive manufacturing approaches into future applications such as artificial muscles, biosensors and bioreactors.

The purpose of this paper is to explain the pointwise Ergodic Theorem and then to apply it to stationary Markov Chains. The Ergodic Theorem is a theorem which shows that the time-averages of a stationary sequence of random variables converge almost surely, and also gives a way to evaluate the limit of these averages. In the setting of Markov chains, the Ergodic Theorem can be used to obtain an important convergence fact about Markov chains.

The advent of RNA-seq technologies provides an unprecedented opportunity to precisely profile the mRNA transcriptome of a specific cell population. It helps reveal the characteristics of the cell under the particular condition such as a disease. It is now possible to discover mRNA transcripts not cataloged in existing database, in addition to assessing the identities and quantities of the known transcripts in a given sample or cell. However, the sequence reads obtained from an RNA-seq experiment is only a short fragment of the original transcript. How to recapitulate the mRNA transcriptome from short RNA-seq reads remains a challenging problem. We have proposed two methods directly addressing this challenge. First, we developed a novel method MultiSplice to accurately estimate the abundance of the well-annotated transcripts. Driven by the desire of detecting novel isoforms, a max-flow-min-cost algorithm named Astroid is designed for simultaneously discovering the presence and quantities of all possible transcripts in the transcriptome. We further extend an \emph{ab initio} pipeline of transcriptome analysis to large-scale dataset which may contain hundreds of samples. The effectiveness of proposed methods has been supported by a series of simulation studies, and their application on real datasets suggesting a promising opportunity in reconstructing mRNA transcriptome which is critical for revealing variations among cells (e.g. disease vs. normal).

Lab-grown tissues have tremendous potential to accelerate drug discovery and identify some of the underlying mechanisms behind diseases. The native extracellular matrix (ECM) of tissues is a complex, hierarchical fibrous protein structure with delicate mechanical properties that guides tissue assembly and regeneration. Existing biomaterial fabrication techniques struggle to simultaneously attain: micro/nano-scale fibril feature resolution, low bulk stiffness and the 3D organisation crucially provided by the ECM without comprising cell motility. This work utilises 3D printing and low voltage electrospinning patterning synergistically to address these conflicting engineering challenges and act as a minimalist guide for 3D cell growth. A version of low voltage electrospinning patterning was adapted as a sequential process on a modified 3D printer. Applied voltage and 3D printed geometry can modulate the suspended behaviour of electrospun fibres that span between 3D printed support pillars, a parametric study characterised threshold conditions and established a predictive model for patterning suspended fibres. The accuracy with which suspended fibres followed the in-plane tool path was also assessed. Scanning Electron Microscopy imaging measured fibre diameters 1-5 μm and mechanical testing examines the properties for a given layer of dry fibres. The configuration demonstrated unique patterning of stacked suspended fibre layers in multiple orientations. Tissue scaffolding applications were explored in 2D and 3D. In 2D, gelatin fibres were patterned as a topographic cue to direct mesenchymal stem cells towards the osteogenic lineage. For 3D cell culture, the use of suspended fibre devices was investigated to improve the efficiency of cerebral organoid assembly. Pursuing these applications led to further refinement of the fibre fabrication technique and the development of targeted cell seeding strategies on suspended fibre structures. Glioblastoma cell aggregates were cultured on suspended fibre devices. Fibres guided the outgrowth of cancer cells from the aggregates, mimicking the topography of white matter tracts that assist migration in vivo. Cells assemble into dense (~200 μm depth) tissue structures with necrotic cores, that can remodel the fibre network yet are guided by the underlying fibre organisation. This novel method of patterning suspended microfibres from solution offers several avenues of inquiry to mimic ECM topography and complex material functionality.

Colloidal polyelectrolyte/counterion micro- and nanogels form automatically via ionic crosslinking, upon mixing multivalent counterions solution with dilute polyelectrolyte solution. Despite the broad application and abundant research of these colloidal materials, most of the studies have been based on trial and error without mechanistic understanding of their formation and coagulation process, and, the relation between molecular and colloidal interactions to their final properties (e.g., size distributions and stabilities). The lack of mechanistic understanding leads to poor control, reproducibility and explanation of experimental results. To address this, isothermal titration calorimetry (ITC) was first used to relate the binding thermogram to the formation and coagulation of polyelectrolyte/counterion colloidal particles. Chitosan/TPP micro- and nanogel system was then studied by ITC, dynamic light scattering (DLS), transmission electron microscopy, zeta-potential and stopped-flow light scattering under various conditions (i.e., pH and ionic strength) to probe the mechanism of the micro- and nanogel formation and coagulation, and, to relate the molecular/colloidal interactions to their final properties(e.g., size distributions and stabilities).

The advancements of GPS and GIS technologies provide new opportunities for investigating vehicle trip generation and destination choice at the microscopic level. This research models how land use and road network structure influence non-work, non-home vehicle trip generation and non-work destination choice in the context of trip chains, using the in-vehicle GPS travel data in the Minneapolis-St. Paul Metropolitan Area. This research includes three key parts: modeling non-work vehicle trip generation, modeling non-work, single-destination choice, and modeling non-work, two-destination choice. This research contributes to methodologies in modeling single-destination choice and multiple-destination choice and tests several hypotheses which were not investigated before. In modeling non-work vehicle trip generation, this research identifies the correlation of trips made by the same individual in the trip generation models. To control for this effect, five mixed-effects models are systematically applied: mixed-effects linear model, mixed-effects log-linear model, mixed-effects negative binomial model, and mixed-effects ordered logistic model. The mixed-effects ordered logistic model produces the highest goodness of fit for our data and therefore is recommended. In modeling non-work, single-destination choice, this research proposes a new method to build choice sets which combines survival analysis and random sampling. A systematic comparison of the goodness of fit of models with various choice set sizes is also performed to determine an appropriate choice set size. In modeling non-work, multiple-destination choice, this research proposes and compare three new approaches to build choice sets for two-destination choice in the context of trip chains. The outcomes of these approaches are empirically compared and we recommend the major/minor-destination approach for modeling two-destination choice. The modeling procedure can be expanded to trip chains with more than two destinations. Our empirical findings reveal that: (1) Although accessibility around home is not found to have statistically significant effects on non-work vehicle trips, the diversity of services within 10 to 15 minutes and 15 and 20 minutes from home can help reduce the number of non-work vehicle trips. (2) Accessibility and diversity of services at destinations influence destination choice but they do not exert the same level of impact. The major destination in a trip chain tends to influence the decision more than the minor destination. (3) The more dissimilar the two destinations in a trip chain are, the more attractive the trip chain is. 4) Route-specific network measures such as turn index, speed discontinuity, axis of travel, and trip chains' travel time saving ratio display statistically significant effects on destination choice. Our findings have implications on transportation planning for creating flourishing retail clusters and reducing the amount of vehicle travel.

This investigation was to compare the structural properties, of commercial available single ply paper towels, as they vary in X-Y plane. The focus was to quantify the differences in selected regions of interest, as measured by two analytical methods, non-contact Twin Laser Profilometry (TLP) and synchrotron X-ray micro-computed tomography (XRμCT). The thickness and out-of-plane deformation (OPD) distributions were mapped using a TLP and XRμCT. The XRμCT samples were further used to calculate for the mass center surface. TLP enables web structures to be mapped to a micrometer resolution over 12.5mm regions. Synchrotron X-ray microtomography method provides sub-micrometer resolution of the complete volumetric structure (1.4mm x 1.4mm x thickness (μm)) for the imaged region. The features of individual fibers are revealed by XRμCT method, so that the true surface geometry may be quantified.

Scope and Method of Study: Scaffolds properties play crucial roles in regulating the cell-matrix interaction. To understand how the scaffolds architecture (i.e. pore structures), the presence of cell-binding domain and stiffness affect cell colonization within the scaffolds, 2D and 3D chitosan and poly (lactide-co-glycolide) (PLGA)-chitosan scaffolds without cell-binding domain were synthesized through controlled rate freezing and lyophilization. Cytoskeletal organization, morphology, and proliferation of Human umbilical vein endothelial cells (HUVECs) and mouse embryonic fibroblasts (MEFs) seeded on 2D, 3D chitosan, PLGA-chitosan scaffolds were characterized. Next, gelatin which contains cell-binding domain was blended with chitosan to investigate cell adhesion characteristics through receptor-mediated binding under static and shear conditions. Murine embryonic stem (mES) cells were used as model to study the differentiation process under the influence of substratum and medium components (i.e. cytokines).

Recreating organ-specific microenvironments of the extracellular matrix (ECM) in vitro has been an ongoing challenge in biofabrication. In this study, I present a biofunctional ECM-mimicking protein scaffold with tunable biochemical, mechanical and topographical properties. This scaffold, formed by microfibres, displays three favorable characteristics as a cell culture platform: high-loading of key ECM proteins, single-layered mesh membrane with controllable mesh size, and flexibility for supporting a range of cell culture configurations. Decellularized extracellular matrix (dECM) powder was used to fabricate this protein scaffold, as a close replicate of the chemical composition of physiological ECM. The highest dECM concentration in the solidified protein scaffold was 50 wt%, with gelatin consisting the rest. In practice, a high density of dECM-laden nano- to microfibres was directly patterned on a variety of substrates to form a single layer of mesh membrane, using the low-voltage electrospinning patterning (LEP) method. The smallest fibre diameter was measured at 450 nm, the smallest mesh size of the membrane was below 1 μm, and the thickness of the membrane was estimated to be less than 2 μm. This fabrication method demonstrated a good preservation of the key ECM proteins and growth factors, including collagen IV, laminin, fibronectin, VEGF and b-FGF. The integrated fibrous mesh exhibited robust mechanical properties, with tunable fibril Young’s modulus for over two orders of magnitude in the physiological range (depending on the dECM concentration). Combining this mesh membrane with 3D printing, a cell culture device was constructed. Co-culture of human glomerulus endothelial cells and podocytes was performed on this device, to simulate the blood-to-urine interface in vitro. Good cell attachment and viability were demonstrated, and specific cell differentiation and fibronectin secretion were observed. This dECM-laden protein scaffold sees the potential to be incorporated into a glomerulus-on-chip model, to further improve the physiological relevance of in vitro pathological models.