Abstract:The purpose of this thesis is to study the homogenization of the lake equation. It generates reaction effect and the reaction term is induced by homogenization due to the weak convergence. It can be characterized by L. Tartar's method of oscillating test function. We begin by the homogenization of a Stokes equation perturbed by a drift. By means of constructing a similar homogenized equation we figure out the weak limit of the important term, this leads to a limit equation with an extra zero-order term. We then turn to the homogenization of an anelastic Stokes system arising from the lake equations and prove the existence of the solution of the equation using Lax-Milgram Theorem. The homogenization of the anelastic Stokes equation lays a fundation for the homogenization of the lake equation. We finally give the proof of the existence and uniquess of the solution of lake equation using Faedo-Galerkin method and study the homogenization of the lake equation by constructing a homogenized equation of the test function according to Tartar's method.

Abstract: Supply chain management has an important role to play in business. To keep a company competitive and survive in international competition, it is important to optimise the processes of production, supply and sales. Today, modern operating companies are faced with an ever-increasing amount of information and problems, and the relationships between supply chain participants are becoming increasingly complex. These problems are even more acute in developing countries. This calls for advanced supply chain design to manage the entire operational processes of the business. Especially in the post-epidemic era, business development in many factories has become unpredictable. Transport logistics and costs are difficult to manage and control accurately. The importance of supply chain resilience is becoming more and more evident. The field of supply chain research needs to break out of its old boundaries and seek more diverse development models and supply chain designs. The results of this thesis first reveal how supply chain leadership affects supply chain performance, providing a theoretical basis for building efficient supply chain. After that, it further reveals the relationship between the risk contagion efficiency of the supply chain network and the individual competitiveness. After that, this thesis emphasizes the impact of resource dependence and operational slack on improving supply chain resilience, and further provides guidance and suggestions for designing supply chain structures in the post-COVID era. Finally, the discussion of supplier relationship and innovation performance provides suggestions on how to improve competitiveness and innovation ability when designing supply chain structure. This thesis conducts a rigorous quantitative analysis of the factors that need to be considered in supply chain design from multiple perspectives and provides a sufficient discussion of their impact, which contributes to research in the field of supply chain management. This thesis also provides policy and corporate operational management guidance for practitioners related to supply chains.

Abstract:Computer Supported Cooperative Work (CSCW) is an emerging technology that integrates human collaborative behaviors and computer-aided techniques. Augmented Reality (AR), Virtual Reality (VR), mobile devices, and large interactive technologies can be used to create unique collaborative experiences, especially for spatial problem-solving tasks. Therefore, this dissertation makes exploratory research on techniques for supporting collaborative activities with consideration of users' performance and experience. It aims to explore two research issues in CSCW: spatial workspace co-presence awareness and proxemics issues. Accordingly, four research questions are answered: RQ1 - Can pointing visual cues support collaborative interaction and improve the awareness of social presence? RQ2 - Is it beneficial to share operations between workspaces for collaborators throughout the sense-making process? RQ3 - Do the proxemics influence collaborative behaviors and task performance? How do position arrangements affect performance, engagement, and collaborative behaviors? and RQ4 - How to give full play to the advantages of different platforms? To address these research questions, four user studies have been conducted to investigate the effect of sharing visualized information and position/device arrangements on technique usability, performance, and user experience in collaborative activities. Four platforms, AR, VR, tablet, and large display, are involved as exploratory environments in this dissertation. The results of this dissertation show that sharing visualized information provides benefits for improving spatial/workspace co-presence awareness to enhance collaboration, and proxemics issues indeed have an influence on task performance, user experience, and behavior during collaboration. Specifically, by exploring the effect of pointing visual cues in AR-supported collaboration, the first study answered RQ1 - pointing visual cues can improve social presence to support collaborative interaction. By exploring the effect of shared operations between workspaces and position/device arrangements in VR, mobile device, and large display collaboration, the other three studies answered RQ2 - sharing operations between workspaces is beneficial for sense-making tasks, and RQ3 - proxemics will influence collaborative behaviors and task performance. Finally, RQ4 was answered by deriving a set of guidelines and design implications for different platforms in co-located collaboration to exploit their advantages. Overall, this research offers some insights into how to design effective multi-user interfaces that are supportive of exploratory tasks.

Abstract:Increasing evidence has suggested that RNA modifications regulate many important biological processes. To date, more than 170 types of post-transcriptional RNA modifications have been discovered. With recent advances in sequencing techniques, tens of thousands of modification sites are identified in a typical high-throughput experiment, posing a key challenge to distinguish the functional modified sites from the remaining ‘passenger’ (or ‘silent’) sites. To ensure that the massive epitranscriptome datasets are properly taken advantage of, annotated, and shared, bioinformatics solutions are developed with various focuses. In this thesis, we first described a comparative conservation analysis of the human and mouse m6A epitranscriptome at single-site resolution. A novel scoring framework, ConsRM, was devised to quantitatively measure the degree of conservation of individual m6A sites. ConsRM integrates multiple information sources and a positive-unlabeled learning framework, which integrated genomic and sequence features to trace subtle hints of epitranscriptome layer conservation. With a series of validation experiments in mouse, fly and zebrafish, we showed that ConsRM outperformed well-adopted conservation scores (phastCons and phyloP) in distinguishing the conserved and non-conserved m6A sites. Additionally, the m6A sites with a higher ConsRM score are more likely to be functionally important. To further unveil the functional epitranscriptome, we investigated the potential influence of genetic factors on epitranscriptome disturbance. Recent studies have found close associations between RNA modifications and multiple pathophysiological disorders, the precise identification and large-scale prediction of disease-related modification sites can truly contribute to understanding potential disease mechanisms. Consequently, we developed a computational pipeline to systemically identify RNA modification-associated variants and their affected modification regions, with emphasis on their disease- and trait-associations. Furthermore, we described the next research considering the dynamics of RNA methylome across different tissues by elucidating the tissue-specific impact of the somatic variant on m6A methylation. The TCGA cancer mutations (derived from 27 cancer types) that may lead to the gain or loss of m6A sites in corresponding cancer-originating tissues were systemically evaluated and collected. Token together, the proposed bioinformatics pipelines and databases should serve as useful resources for functional discrimination and annotation of the massive epitranscriptome data, with implications for the potential disease mechanisms functioning through epitranscriptome layer.

Abstract:Since the building sector accounts for more than one third of global carbon emissions, it is imperative that the sector mitigate its emissions to help reach the goal of the COP26 climate conference of achieving a global net zero by mid-century. Building refurbishment (BR) is key to reducing carbon emissions in the building sector by reducing the operational energy consumption of existing buildings instead of demolishing them and building new ones. China is a good example of a country encouraging refurbishment, since it has prioritized BR in its 14th Five-Year Building Energy Efficiency and Green Building Development Plan (2021-2025). Since the number of BR projects in China is therefore likely to significantly increase in the coming years, it is important to evaluate the carbon emissions associated with construction and demolition (C&D) waste to find optimal waste management solutions. However, there are no studies that have considered the carbon emissions of C&D waste management of BR projects from a whole life cycle perspective. This study fills the research gap by developing a novel LCA-SD model, which integrates the features of life cycle assessment (LCA) and system dynamics (SD) to evaluate the carbon emissions of C&D waste management of BR projects through non-linear and dynamic analysis from a whole life cycle perspective. Variables for evaluating the carbon emissions were first identified in four life cycle stages of C&D waste management of BR projects. Causal loop diagrams were then developed to demonstrate the interrelations of the variables in the different life cycle stages, and the novel LCA-SD stock and flow model was formulated based on the causal loop diagrams. The model was validated through a case study of a typical BR project in China. The validated LCA-SD model was used to compare and analyze waste management scenarios for the case study BR project by performing simulations of selected scenarios. The simulation results reveal that the secondary material utilization rate is the most effective independent variable for reducing carbon emissions from C&D waste management of the case BR project, 11.28% of total carbon emissions could be reduced by using 31% of secondary materials to substitute natural raw materials; improving the combustible waste incineration rate to 100% could reduce 6.42% of total carbon emissions; reducing 50% of the on-site waste rate could reduce 1.28% of total carbon emissions; while improving the inert waste recycling rate to 90% could only reduce 1% of total carbon emissions. From the whole life cycle perspective, the refurbishment material stage accounts for the highest carbon emissions, followed by the refurbishment material EOL stage, and the dismantlement stage, the refurbishment construction stage accounts for the least carbon emissions. The findings not only highlight the importance of cradle to cradle life cycle C&D waste management for mitigating carbon emissions from BR projects, but also demonstrate the effectiveness of the novel integrated LCA-SD model as an “experimental laboratory” for BR C&D waste management decision makers to conduct “what-if” dynamic simulation analysis for various scenarios before embarking on a project.

Abstract:Conventional methods for flood frequency analysis and prediction are made on the basis of the stationary assumption that assumes the probability of flood flows is unchanging over time. However, observed and potential non-stationarity in floods can result in costly overdesign or dangerous under-design. Despite increasing attention, there is no consensus on a methodology for flood frequency analysis in nonstationary conditions in the literature. In addition, river flows are affected by various factors, and therefore the dynamics in their associated time series are complicated and have nonlinear and nonstationary behaviours, which also pose a challenge to traditional methods for flow prediction. This thesis focuses on flood frequency analysis and river flow prediction in nonstationary environments aiming to provide new insights into flood-generating mechanisms and improve flood prediction in a changing environment. First, the thesis presents a state-of-the-art review of the challenges and advances in detection, attribution, and management of the non-stationarity in flood series. Second, peak river flow series from 158 gauging stations in the UK are analysed using both stationary and nonstationary models based on the generalised additive models for location, scale and shape (GAMLSS) framework. The results show that very few stations are found to show temporal trends, which suggests that time-varying non-stationary flood models may not always be suitable, and the use of physically-based covariates can provide better-fitting models. In particular, the non-stationary models using rainfall-related covariates display the best performance for the vast majority of stations, thus suggesting that the variability of the rainfall regime remains the dominant driver for changes in flooding. Two atmospheric circulation patterns, namely the North Atlantic Oscillation (NAO) and the East Atlantic (EA) Pattern, were found to be closely related to the non-stationarity of flooding at a number of stations. For the majority of stations, the use of multiple covariates can provide a better non-stationary model than the use of a single covariate. In the next strand, to identify a suitable and effective model for better performing nonstationary flood frequency analysis (NFFA), different modelling techniques, including linear, nonlinear, parametric and nonparametric regressions are assessed. For both time-varying and precipitation-informed NFFA, rejection rates for linear and cubic polynomial models are the highest. The models with the fewest rejections are fractional polynomial models followed by cubic spline models. Because of the flexibility, parsimony, and user-friendly features, fractional polynomial models could be a potential alternative for modelling the non-stationary behaviour of flood series. In the last technical part, a fresh hybrid forecasting model, NeuralProphet (NP), has been applied to river flow prediction in four upland catchments in Scotland. It is found that the NP model can accurately capture the complicated nonlinear features of monthly peak flow series. The use of deep learning techniques in NP models can improve the prediction accuracy significantly and the model training is very fast. Most importantly, it is desirable for forecasting tools to be not only accurate but also accessible for decision-makers. The NP model can be decomposed into components which are individually interpretable. This allows for the prediction to be reviewed by hydrologists or decision-makers who may make adjustments when appropriate. This thesis increases the understanding of nonstationary flood behaviour and provides new insights into managing future risks arising from nonstationary water extremes. Also, the findings of this thesis can provide practical recommendations to hydrologists and engineers when implementing NFFA, and the results of NFFA have potential implications for the design and/or economic justification of measures to reduce flood risk.

Abstract:Image inpainting aims to restore deteriorated images with plausible and relevant contents, which has wide applications in many computer vision tasks. Conventional methods based on pixels propagation or patch replacement often fail in complicated scenarios with ambiguous artifacts or blurry fillings. In recent years, deep generative models constructed with autoencoder and generative adversarial networks have been exploited extensively and shown impressive results. Deep generative models are more advanced in understanding image global information through large-scale data than traditional models. However, those deep leaning-based approaches still have challenges in terms of bad generation quality under complex scenarios (i.e., artifacts and blur). Besides, they are unable to fit into real-world applications efficiently. This dissertation summarizes the proposed work on deep learning-based image inpainting to address these challenges theoretically and practically. Theoretically, this thesis analyzes and proves that multi-scale architecture can isolate the learning of structures and textures, thus generating high-quality inpainting results. The proposed model includes a pyramid of generators, each responsible for progressively capturing the image global structure and local details. Meanwhile, this thesis has studied using the feature fusion technique to address the insufficient utilization of current encoder-decoder features and further alleviate the low-quality synthesized patches. From a practical perspective, this thesis is committed to building user-oriented interactive inpainting systems for real-life uses. To this end, an efficient object removal pipeline integrated with object segmentation technique and an attention-based structural guided module is proposed. Overall, this thesis shows theoretically and practically deep generative model is a powerful tool to address these challenges. Nonetheless, the last part presents the novel contributions of this thesis and the outlook for future work.

Abstract:With the development of society, both industry and academia draw increasing attention to multimedia systems, which handle image/video data, audio data, and text data comprehensively and simultaneously. In this thesis, we mainly focus on multi-modality data understanding, combining the two subjects of Computer Vision (CV) and Natural Language Processing (NLP). Such a task is widely used in many real-world scenarios, including criminal search with language descriptions by the witness, robotic navigation with language instruction in the smart industry, terrorist tracking, missing person identification, and so on. However, such a multi-modality system still faces many challenges, limiting its performance and ability in real-life situations, including the domain gap between the modalities of vision and language, the request for high-quality datasets, and so on. Therefore, to better analyze and handle these challenges, this thesis focuses on the two fundamental tasks, including matching and segmentation. Image-Text Matching (ITM) aims to retrieve the texts (images) that describe the most relevant contents for a given image (text) query. Due to the semantic gap between the linguistic and visual domains, aligning and comparing feature representations for languages and images are still challenging. To overcome this limitation, we propose a new framework for the image-text matching task, which uses an auxiliary captioning step to enhance the image feature, where the image feature is fused with the text feature of the captioning output. As the downstream application of ITM, the language-person search is one of the specific cases where language descriptions are provided to retrieve person images, which also suffers the domain gap between linguistic and visual data. To handle this problem, we propose a transformer-based language-person search matching framework with matching conducted between words and image regions for better image-text interaction. However, collecting a large amount of training data is neither cheap nor reliable using human annotations. We further study the one-shot person Re-ID (re-identification) task aiming to match people by offering one labeled reference image for each person, where previous methods request a large number of ground-truth labels. We propose progressive sample mining and representation learning to fit the limited labels for the one-shot Re-ID task better. Referring Expression Segmentation (RES) aims to localize and segment the target according to the given language expression. Existing methods jointly consider the localization and segmentation steps, which rely on the fused visual and linguistic features for both steps. We argue that the conflict between the purpose of finding the object and generating the mask limits the RES performance. To solve this problem, we propose a parallel position-kernel-segmentation pipeline to better isolate then interact with the localization and segmentation steps. In our pipeline, linguistic information will not directly contaminate the visual feature for segmentation. Specifically, the localization step localizes the target object in the image based on the referring expression, then the visual kernel obtained from the localization step guides the segmentation step. This pipeline also enables us to train RES in a weakly-supervised way, where the pixel-level segmentation labels are replaced by click annotations on center and corner points. The position head is fully-supervised trained with the click annotations as supervision, and the segmentation head is trained with weakly-supervised segmentation losses. This thesis focus on the key limitations of the multimedia system, where the experiments prove that the proposed frameworks are effective for specific tasks. The experiments are easy to reproduce with clear details, and source codes are provided for future works aiming at these tasks.

Abstract:Video surveillance outputs different portrait information of scenes such as crime investigation, security system, automatic driving system, and environmental monitoring. Recently, deep learning based video surveillance is also an essential topic in computer vision. The specific tasks include object tracking, video object segmentation, salient object detection, and video salient object detection. Thus, this thesis studies salient object detection and segmentation in video surveillance, mainly on video object segmentation and salient object detection. In video object segmentation, we study the case of given the first frame's mask and try to design a network that can adapt to different object appearance variations. Therefore, this thesis proposes a framework based on the non-local attention mechanism to localize and segment the target object in the current frame, referring to both the first frame with its given mask and the previous frame with its predicted mask. Our approach can achieve 86.5$% IoU on DAVIS-2016 and 72.2% IoU on DAVIS-2017, with a speed of 0.11s per frame. Then for salient object detection, this thesis focuses on scribble annotations. However, scribbles fail to contain enough integral appearance information. To solve this problem. A local saliency coherence loss is proposed to assist partial cross-entropy loss and thereby help the network learn more complete object information. Further, A self-consist mechanism is designed to help the network not sensitive to different input scales. Our method can achieve comparable results compared with fully supervised methods. Our method achieves a new state-of-the-art performance on six benchmarks (e.g. for the ECSSD dataset: Fβ = 0.8995, Eξ = 0.9079 and MAE = 0.0489). Lastly, co-salient object detection is also studied. Recent methods explore both intra- and inter-image consistency through an attention mechanism. We find that existing attention mechanisms can only focus on limited related pixels. Thus, we propose a new framework with a self-contrastive loss to mine more related pixels to obtain comprehensive features. Our method obtains 0.598 for maximum F-measure for COCA. In this way, the tasks in this thesis are well handled and our methods can serve as new baselines for future works.

Abstract:With the increasing demand and scale of power networks, the outages caused by cascading failures have catastrophic consequences on grid stability and reliability, severely impacting public life and socio-economic costs. Furthermore, complex networks (CN) theory is widely studied and applied to reveal the network's structure and further analyze the characteristics of the network, such as social, economic, power, transportation, and biological networks. Community detection is an important technique to discover the structure of a network by dividing a large-scale network into subgroups. However, the functionality of networks also can notcannot be neglected when understanding community structure. Community detection methods should consider the different functional and physical rules of different network systems to investigate network characteristics. Furthermore, few people have systematically studied the role of community structure in power grid security analysis. In this context, this thesis focuses on the power grid structural analysis based on the complex networks theory. First, the flow-based community in engineering networks is studied. The flow-based communities are defined and detected based on a new index and corresponding functional modularity to understand community structures in engineering networks. A novel index is proposed to evaluate partitioning results by upgrading the conventional modularity. Moreover, three different definitions of the proposed index are introduced to community detection in three different categories of networks. Furthermore, a metric for a specific community is proposed so that communities of different scales in different networks can be quantitatively compared.  Second, the study of flow-based communities is shifted to power networks. A functional community is defined with two meanings. New weight is proposed to reflect the two specific meanings in power supply communities combined with the structural and functional characteristics of power grids. Then, this work proposes a new index as a benchmark to evaluate any partitioning of power grids. Furthermore, the conventional Newman fast algorithm was modified based on power supply modularity. Third, inspired by the characteristics of the 2012 India blackouts, a new community structure is defined as the structure of several communities in which the transmission of physical elements between communities is significantly stronger than interaction within the same community. This new structure captures the critical characteristics of operating states leading to cascading failures.  Finally, based on the proposed community structure, this work proposes two strategies for power grid security from the perspectives of protector and attacker respectively. Considering the risk of cascading failures and the cost of power network operation, a new optimal power dispatch method is proposed. The simulation results on the IEEE 118-bus system demonstrate that this approach can mitigate the threats of cascading failures while maintaining the cost economy. As a typical Cyber-Physical System (CPS), a modern power system may suffer from increasing risks of Coordinated Cyber-Physical Attacks (CCPA). Then, from the perspective of the attacker, a novel cyber-physical attack mode is proposed based on false command injection where cyber-attacks could lead to following physical attacks by inveigling system operations to vulnerable states.

Abstract:Textiles have always been an important carrier for people to express their emotions. Wearable technology brings the power of computing closer to the body, which has led to the rising focus of bodily emotion-related interactive textiles design. Body motion-engaged affective design for emotion regulation also become a trend in games and interactive art, but the advantages of interactive textiles in this direction have not been thoroughly studied. This research aims to answer the overarching question: What are the key factors influencing body-centred interactive textiles design for emotion regulation? First, we conducted a comprehensive review of the literature on emotion-related interactive textiles design and combed the current state of body-centred interactive textiles and the opportunities that exist. From this, we discussed its research directions in (1) exploring the materiality of interactive textiles, (2) integrating body motion into interaction, and (3) creating sensory stimuli with body-centred interactive textiles. Then, we investigated the research question following the Research through Design approach and verified the feasibility of regulating emotion through body-centred interactive textiles. Five sub-research questions were identified to concrete the research direction. In the design research and practice, first, we identified effective gross movement-based interactive textiles with multi-sensory feedback mechanisms for emotion regulation and created the E-motionWear prototype for participants to test in the lab and real life. Then, we presented GesFabri, a collection of five interactive textile interfaces with distinct textures. They were created to investigate the intuitive interaction gestures of textile texture and the emotional effects of the fine movement-based interfaces with four sensory feedback mechanisms. Last, we developed iPillowPal, an affective movement-based interaction pillow for emotional communication and engagement between long-distance relationship (LDR) partners. We conducted a user study, lab study, and field study to achieve this goal. The results revealed that (1) wearing the gross movement-based interactive textiles positively impacted the users’ immediate emotion regulation, and users presented a more positive attitude towards their work. (2) both fine movement-based interaction on textiles and the feedback mechanism influenced user emotions, also highlighted the potential of materiality in designing interactive textiles, (3) affective movements vary on the scenario, serve a specific emotional purpose, and generate a particular emotional response. The affective movement-based interactive textiles shortened the emotional distance of LDR couples and improved their emotional state. Multi experimental results show that several emotion regulation strategies can be applied to body-centred interactive textiles design. Based on the findings, we summarized our research through the lenses of body movement, feedback mechanism, textile materiality, and emotion types, then reflected on the research approach and methods, design implication, and proposed future work directions. This study offers a new perspective on the potential of emotion regulation with interactive textiles and will contribute to a deeper understanding of body-centred affective design.

Abstract:Extant studies have reached an agreement that family firm internationalization is unique, not only because of the complex trade-off between economic and non-economic wealth involved in the decision-making of such firms but also because of the opportunity they present to examine and enrich theories in the field of management and international business. Motivated by the uniqueness of family firm internationalization, this thesis focuses on the impact of the unique governance features of family firms on their international activities, and it presents three empirical studies to address the research gaps identified by a systematic literature review. This thesis investigates why and how family involvement influences international resource allocation. In family firms where socioemotional wealth (SEW) preservation has a higher priority than financial performance, family firms are suggested to treat different types of resources asymmetrically. This thesis empirically supports the view that family firms are less likely than non-family firms to allocate financial slack to foreign markets but are more likely to allocate human resource (HR) slack to such markets. Moreover, the thesis also asserts that when family firms receive negative performance feedback, their priority in decision-making will be altered from SEW preservation to bolstering financial performance. Leveraging the psychological ownership theory, the thesis finds that internally promoted non-family CEOs tend to engage in internationalization because they have stronger psychological ownership with the firm through their knowledge, control and devotion to the firm. However, the involvement of the incoming family generation and institutional investors reduces the internally promoted non-family CEOs’ tendency to conduct internationalization. The thesis also pays attention to tax haven internationalization by analysing how family owners balance the costs and benefit associated with SEW preservation when family firms establish tax haven subsidiaries. The positive relationship found between family ownership and the likelihood of establishing tax haven subsidiaries reveals a dark side of SEW preservation that can also promote family firms’ unethical behaviours. In sum, this thesis examines the impact of family and non-family involvement on internationalization and contributes to family firm literature and international businessliterature in a number of ways. In particular, this thesis contributes to the family firm literature by explaining the asymmetric treatments that family firms hold towards different resources, by examining the boundary conditions of prioritizing SEW preservation, by revealing the dark side of SEW preservation, and by distinguishing between different types of non-family CEO. Meanwhile, the thesis explores the impact of slack resource on internationalization from the perspective of strategic intent and takes a balanced view to analyse the costs and benefits associated with tax haven internationalization, thus contributing to the international business literature.

Abstract:The large and diverse supra-family of RNA-binding proteins (RBPs) orchestrates the fate of cellular RNAs starting from biosynthesis to decay either via direct interaction or via ternary formation of ribonucleoprotein (RNP) complexes. Interestingly, recent technical advances allowed for the discovery of hundreds of previously unrecognized RBPs in human cells, with many of such factors surprisingly containing unconventional, disordered or yet uncharacterized RNA-interacting modules. A substantial number of these RNA-omics studies also revealed the unique human DEK oncoprotein, a conserved factor known to interact with cellular DNA and chromatin, as an RBP. Yet, the precise RNA interaction domain in DEK remained elusive. A recent study from our research group identified the central region of DEK (amino acids 187-270) as a prominent RNA-interaction domain by employing a newly developed biochemical method, namely Bacterial Growth Inhibition Screen (BGIS). This domain in DEK appears to be intrinsically disordered and shows no resemblance to any yet known or characterized RNA-interaction domains. Based on these findings, the major aim of this dissertation was to elucidate the biological role of the DEK-RNA interplay in human cells. To tackle this task, an “RNA-binding null” mutant, termed RBN#3c, within the lysine-rich region of DEK (amino acids 187-270) was successfully created by employing BGIS in combination with random mutagenesis. A cross-linking and immunoprecipitation (CLIP) approach was chosen to identify distinct DEK-interacting RNA species. Indeed, using lentivirally-engineered human cell lines inducibly expressing eGFP fusions of DEK full length and RBN#3c for CLIP-sequencing (CLIP-seq) approaches, various distinct RNA species, encompassing mRNAs, long non-coding RNAs, snoRNAs, snRNAs, miRNAs as well as rRNAs, were identified. Bioinformatics analyses of RNA sequences cross-linked to DEK revealed that the central RNA-binding region of DEK may predominantly function in pre-mRNA splicing, based on the “purine-rich” identity of RNA transcripts captured. Surprisingly, the C-terminal portion of DEK, also able to interact with RNA, preferentially binds to the poly(A)-tail of mRNA, suggesting distinct roles of these two RNA-interaction domains in DEK. Importantly, further detailed bioinformatics analyses of the global RNA target datasets revealed strong overrepresentation of a particular cellular pathway – ribosome biogenesis, which was somewhat unexpected given the previously known functions of DEK. Indeed, subsequent experimental cytoplasmic ribosome profiling analyses in DEK-depleted cells strongly supported this notion. Specifically, a DEK-dependent pronounced impairment of the 60S large ribosomal subunit along with substantially reduced global translation efficiency in cells was observed. Importantly, this pronounced deficiency could be rescued by re-expression of Cterminal DEK fragments, yet not N-terminal fragments or the RBN#3c mutant. Detailed analyses of the ribosome biogenesis pathway using a series of state-of-the-art techniques, including SUrface SEnsing of Translation (SUnSET) assay, Fluorescence In Situ Hybridization (FISH) analyses, Pre-ribosome Sequential Extraction (PSE) assay, and others, revealed implications for DEK at multiple stages along this complex pathway: rRNA transcription, pre-rRNA processing and transport of pre-ribosomal particles. Most importantly, detailed cellular experiments uncovered the existence of a previously unknown role for DEK in activation of p53 (TP53) in response to cellular stress via regulation of the localisation of the 5S ribonucleoprotein (RNP). Specifically, our data demonstrated that depletion of DEK leads to ribosomal stress, resulting in disruption of the nucleolus and the release of the 5S RNP complex to a free pool, which then interacts with MDM2. The released ribonucleo-complexes are then redirected to interact with MDM2 and block the E3 ubiquitin ligase activity of MDM2 towards p53, thus eliciting p53 activation, which is core to the so-called Impaired Ribosome Biogenesis Checkpoint (IRBC) pathway. Moreover, we defined a direct role for DEK in regulating 5S RNP localisation via specific binding to 5S rRNA, which enables for new perspectives of the involvement of an oncoprotein in the IRBC pathway, proposing crosstalk between cellular responses to oncogenic and ribotoxic stress. Collectively, this thesis, for the first time, at least to our knowledge, examined the RNA interactome of human DEK under physiological conditions. The outcome revealed an unexpected aspect of cellular DEK biology: regulation of the ribosome biogenesis pathway at multiple stages. We now identify DEK as a potential regulator of the central tumour suppressor p53 in p53-intact cells via the IRBC checkpoint - a prominent anti-cancer barrier. Our findings not only expand the knowledge about this notoriously multifunctional and unique oncogene, yet they also reveal clinically relevant functions which may allow paving the way to new approaches that target DEK for tumour intervention strategies.

Abstract:Virtual Learning Environments (VLEs), as the critical educational technology, have the potential to enable new learning opportunities (e.g., personalized adaptive learning and seamless mobile learning) and promote educational innovations for sustainable educational change in Higher Education (HE). While the research on VLEs and technology-enhanced learning in HE has been promising, the adoption of VLEs and the diffusion of educational innovations are not as widespread as expected, and the mechanism is unclear. Additionally, most the Higher Education Institutions (HEIs) have, until the recent COVID-19 disruption, been somewhat cautious about the potential educational reform. Whether the emergency educational transition is temporary or sustainable remains an open question. This research programme investigates how technology promotes educational change through six sub-studies of VLEs in HE. Five studies have been published as journal papers, while one is under review. Firstly, a systematic literature review was conducted to analyse the recent studies of VLE adoption from 2001 to 2020. Two-factor categories - institutional and individual were synthesized from 290 factors identified from findings of 145 studies across 42 countries and regions. Consequently, knowledge gaps of the institutional and individual factors and mechanisms were further investigated by conducting five studies from multiple perspectives. Specifically, four empirical studies examined three key aspects (institutional normative facilitating, institutional cognitive-cultural influence, and individual cognitions) of the VLE adoption and educational innovation institutionalisation in a Sino-British international university in China. The four studies employed various research methods (qualitative, quantitative, and mixed) to investigate technology promoted educational changes in different institutional stages at the individual and organisational levels. Finally, a conceptual study was conducted to reconceptualise the digital learning ecology model based on the existing literature and empirical findings at an institutional field level, a social arena in which individuals and organizations share a common meaning system (Scott, 2004). The main results revealed that VLEs could promote sustainable educational change through the technology-human interactions that are directed by the two-dimensional meaning-making process: collective cognitive consensus (i.e., national culture and learning community) and individual cognitive divergence (i.e., perceived pedagogical value, perceived self-efficacy, and perceptions of justice). This research programme contributes to the literature on Education, Information Technology, Psychology and Sociology by extending people’s understanding of the existing theories and models (e.g., Unified Theory of Acceptance and Use of Technology, Institutionalisation Model, Social Learning Theory, Organizational Justice Theory and Equity Theory, HeXie Education Model) through the theory development and reconceptualisation. This research programme provides theoretical and practical implications to address the grand challenges in HE (i.e., success in technology adoption, widespread innovations, and sustainable educational change). The research findings suggested that educational policy makers and practitioners should include teachers and students as the co-creators of the future digital learning ecosystem, provide continuous teacher professional development in technological and pedagogical skills and knowledge, and develop student competence in self-directed learning and digital resilience. Educators, learners, and researchers should utilise the findings and supporting methods to develop innovative learning and teaching approaches. Future research is needed to test the theoretical models in other educational contexts (e.g., K12 and vocational) and geographies with larger samples to enhance global development.

Abstract:To overcome the practical constraint that the test data should be in the same feature space and follow the same distribution as the training data, transfer learning is proposed to achieve the specific task on a target domain by transferring the task-relevant knowledge from a different source domain. Zero-shot learning, as a sub-field of transfer learning, aims to achieve the classification of target classes without corresponding labelled training samples. It was proposed to imitate the efficient human learning ability that constructs concepts of unknown classes based on relevant descriptions and learned categorical knowledge. To solve the challenging task where samples corresponding to target classes are invalid during training, researchers proposed approaches rely on two main ideas, embedding and generation, respectively. Since the generative methods synthesize pseudo samples for unseen classes based on the corresponding semantic attributes, the followed training process for the classifier might be regarded as breaking the strict target unknown principle. In the inductive scenario where only seen classes are available during training, embedding methods draw more focus as they could learn a target space only depending on the training classes to achieve classification via settled or learned metrics.  With the methods steadily improved, different problem settings, and diverse experimental setups have emerged, the effectiveness of the proposed methods could be inappropriately evaluated. Thereby, in this dissertation, we first provide a comprehensive survey on zero-shot image classification to provide a thorough introduction to this field. Particularly, we have examined three implementation details that can boost the performance of zero-shot learning, i.e. whether the backbone structure has been modified, whether fine-tuning has been conducted, and whether additional knowledge has been used. By annotating these experimental details, we have collected a more careful comparison between various zero-shot methodologies. The rest part of the dissertation summarizes our work which focuses on improving the metric for the embedding methods under the inductive zero-shot learning scenario. Due to the absence of the labelled target samples in the training stage, the learned embedding space or metrics is easily over-fitted for those seen classes thus leading to the model incorrectly predicting the unseen class as one of those in training when the test label space covered both the seen and unseen classes. To alleviate such an over-fitting problem, we proposed a self-focus mechanism for a ridge regression based method. The proposed mechanism takes the embedded semantic attribute vector as input to produce focus ratios for the dimensions in the embedding space. When these ratios are used for constructing the optimization loss, the correlations between the location and the importance of each dimension are considered. Thus the learned embedding space will be more generalized for classification. However, this mechanism can not be flexibly applied to the methods with learnable metrics. We then, proposed two adversarial frameworks on the sample and parameter spaces, respectively, for the relation network based methods. The designed frameworks help train a robust model on seen data and enhance the sensitivity of unseen classes through adversarial perturbations. As a result, the learned model returns high responses to unseen classes while not affecting the recognition of seen classes due to the robustness.

Abstract:The rapid economic and social development caused an increasing demand for industrial and civil construction. However, with the growing number of construction projects, the shortage of natural materials for mixing concrete has reached a crisis. There have been increasing studies focusing on introducing recycled wastes such as fly ash (FA) used for replacing cement and recycled coarse aggregates (RCA) used for replacing natural gravel with concrete. However, the inadequate mechanical performance and durability of FA concrete and RCA concrete (RCAC) have limited their widespread use. Biomineralisation has been evidenced as a helpful way to improve the mechanical performance and durability of concrete. Ureolytic bacteria are common strains used to induce biomineralisation. This research has preliminarily adopted three ureolytic bacteria strains for application in concrete. Adding bacterial solution into mixing water and pretreating RCA with bacteria solution and using these bio-treated RCA as the coarse aggregate concrete are the two approaches used in this project for applying the biomineralisation method to concrete. The results have shown that water absorption of RCA decreased by 17.3%, and the apparent density increased by 2.4% after the treatment. Meanwhile, applying the same type of bacteria with nutrition and extra calcium ions into fresh concrete, the compressive strength of FA concrete improved by 30%. The mechanical properties of concrete with 40% FA and 30% treated RCA enhanced by Sp bacterial solution can be close to normal concrete with the same mixing design. However, ammonia (NH3) was generated unavoidably with an unpleasant smell. As part of the research focus, the use of denitrification bacteria to induce biomineralisation has also been investigated. Like the ureolytic bacterial approach, this approach also exhibited satisfactory enhancing effects and was researched to enhance RCAC. Pre-treating the RCA by soaking in bacteria solution and mixing bacterial solution were proposed to apply to RCAC together, which was abbreviated as BTRCAC (bacterial treated RCAC). For BTRCAC, 30.3% and 19.2% improvements were recorded for compressive strength and tensile splitting strength, respectively. The calcium carbonate (CC) induced by the bacteria filled the open pores inside the concrete, which contributed to a 33.0% decrease in water absorption, impeding the absorption of outside water and reducing the swelling caused by water freezing. Bacteria bond the ITZs, resulting in higher cohesion between aggregates and matrix as well as enhancing the resistance to the pressure caused by freeze-thaw cycles. BTRCAC can withstand 75 more freeze-thaw cycles than RCAC. LCA analysis found that environmental impact (EI) contributions created by bacterial solution production were negligible. Though the EI caused by using nutrition and calcium sources is high, however using recycled wastes decreases EI and the biomineralisation method compensated for the inadequate performance caused by using the recycled wastes. Hence, compared with ordinary concrete, bio-based FA concrete with 100% RCA decreased  27.0% of Global Warming Potential (counted by kg CO2 eq). Moreover, the normalised EI value, which was obtained by weighted summing multi-EI categories including Global Warming Potential, Acidi?cation Potential, Photochemical Ozone Formation Potential, Particulate matter, and Ionizing radiation according to EU Environmental Footprint 3.0, of bio-based recycled waste concrete was reduced by a maximum of 25.0%. Furthermore, if the nutrition and calcium ions can be replaced by the contaminant in wastewater, the EI of the bio-based recycled waste concrete can be further significantly decreased. Therefore, the bio-based recycled waste concrete w is a sustainable way to replace the traditional concrete.

Abstract:Crowdsourcing has become an international phenomenon attracting businesses and a crowd workforce across the globe. China, being one of the world’s most populous countries, has a rapidly growing digital economy that now supplies a substantial workforce to crowdsourcing platforms. However, not only is there limited research on the work experiences and practices of Chinese crowdworkers, but they generally overlook issues pertaining to an emerging type of crowd workforce known as “crowdfarm” - that of organizations taking and undertaking crowdwork as part of their formal businesses. The lack of understanding about the involved digital workforce has been identified as an obstacle to the development and application of crowdsourcing as a disruptive value creation model utilizing the resources of human intelligence.  Therefore, considerable potential exists in the Chinese crowdsourcing context for HCI and CSCW studies to contribute to the alleviation of this issue. This thesis explores the job demands, resources, crowdwork experiences and platform commitment of the general Chinese crowdworkers, compares the work experiences of crowdfarm workers and solo crowdworkers, and examines the work practice of crowdfarms as well as their interplays with solo crowdworkers, requestors, and crowdsourcing platforms. In order to explore the aforementioned, first, based on a framework of well-established approaches, namely the Job Demands-Resources model, the Work Design Questionnaire, the Oldenburg Burnout Inventory, the Utrecht Work Engagement Scale, and the Organizational Commitment Questionnaire, we systematically study the work experiences of 289 crowdworkers who work for ZBJ.com - the most popular Chinese crowdsourcing platform. Our study examines these crowdworker experiences along four dimensions: (1) crowdsourcing job demands, (2) job resources available to the workers, (3) crowdwork experiences, and (4) platform commitment. Our results indicate significant differences across the four dimensions based on crowdworkers’ gender, education, income, job nature, and health condition. Further, they illustrate that different crowdworkers have different needs and threshold of demands and resources and that this plays a significant role in terms of moderating the crowdwork experience and platform commitment. Overall, this work part sheds light to the work experiences of the general Chinese crowdworkers and at the same time contributes to furthering understandings related to the work experiences of crowdworkers. Next, drawing on a study that involves 48 participants, our research explores, compares and contrasts the work experiences of solo crowdworkers to those of crowdfarm workers. Our ?ndings illustrate that the work experiences and context of the solo workers and crowdfarm workers are substantially different, with regards to all of the investigated seven aspects, namely (1) work environment, (2) tasks, (3) motivation and attitudes, (4) rewards, (5) reputation, (6) crowdwork satisfaction, and (7) work/life balance. This part of the work contributes to furthering the understanding of the work experiences of two different types of crowdworkers in China.  Finally,  we have extended our study of typical solo crowdworker practices to include crowdfarms. We report on interviews of people who work in 53 crowdfarms on the ZBJ platform. We describe how crowdfarms procure jobs, carry out macrotasks and microtasks, manage their reputation, and employ different management practices to motivate crowdworkers and customers. The results also reveal the crowdfarms’ interplay with solo crowdworkers, requestors and crowdsourcing platforms.  Overall, this work provides one of the first systematic investigations of the work experience and practice of digital labors in the Chinese crowdsourcing context, addressing the relevant gaps in the current literature. At the same time, by identifying and studying an emerging crowdsourcing workforce - crowdfarm - in the changing landscape of crowdsourcing in China, our work also provides a new direction and topic for researchers in the field of HCI/CSCW. We hope our work stimulates others to join in research and discussion of the potential impact of such evolution on the gig economy and the well-being of the tens of millions of people now engaged in crowdsourced work in a broader context.

Abstract:As an important application scenario of deep learning, Natural Language Processing (NLP) is receiving more and more attention and developing rapidly. Learning representation for words or documents via neural networks is gradually replacing feature engineering in almost all text-related applications. On the other hand, how to decode these representations or encodings is also very vital for sequence-to-sequence text generation tasks such as Neural Abstractive Summarization (NAS), Neural Machine Translation (NMT), etc. Towards a more comprehensive representation and decoding strategy, this dissertation explores several global perspectives that previous studies ignored. We treat global as a relative concept that indicates higher-level knowledge conducive to enriching representation or improving decoding. However, its specific definition may vary in different tasks.  In text representation or encoding, global refers to relatively higher-level context information. There usually are three natural contextual relationships for mapping words or documents into latent space, namely (1) co-occurrence relationships between words, (2) coherence relationships between sentences, and (3) subordinate relationships between documents/sentences and their words. Beyond these naturally occurring contexts, there are possibly hidden context relationships between dependent documents from the perspective of the whole corpus (i.e., the global perspective). Although we often assume that documents in a corpus are independent of each other, the assumption may not be valid for some corpora like news corpora, since events reported by news documents interact in the real world. To capture the global-contextual information, we construct a news network for the whole corpus to model the latent relationships between news. A network embedding algorithm is then designed to produce news representations based on the above-mentioned subordinate relationship and news dependency. Besides, such a cross-document relationship plays a vital role in some specific tasks which need to represent or encode a cluster of multiple documents, e.g., Multi-document Summarization (MDS). Some studies concatenate all documents as a flat sequence, which is detrimental to modeling the cross-document and long-term dependency. To alleviate the two problems, we design a Parallel Hierarchical Transformer (PHT), whose local and global attention mechanisms can simultaneously capture cross-token and cross-document relationships. On the other hand, global in text decoding refers to a higher-level optimum, i.e., the global optimum relative to the local optimum. Under the fact that the neural text generator is almost impossible to generate the whole sentence at once, the heuristic algorithm -- beam search has been the natural choice for text decoding. Inevitably, beam search often gets stuck of local optimum as it decodes word-by-word. Although global optimum is hard to touch directly, it is feasible to conduct a one-shot prediction of how the global optimal hypothesis attends to the source tokens. A global scoring mechanism is then proposed to evaluate generated sentences at each step based on the predicted global attention distribution, thus calibrating beam search stepwise to return a hypothesis that can assign attention distribution to the source in a more-near global optimal manner. Decoding with global awareness improves the local optimum problem to enhance the generation quality significantly, and it can be developed and used in various text generation fields. 

Abstract:Thin-film transistors (TFTs), serving as the core components for the applications of the active matrix for liquid crystal displays (AMLCDs) and the active matrix for organic light emitting diodes (AMOLEDs), have been being intensively researched all over the world. For the past decades, in order to meet the display application requirements of high resolution, large screen size, and low power consumption, the metal oxides (MOs) semiconductors have been proposed and widely investigated for the fabrication of high-performance TFTs. Compared with the traditional TFTs based on the amorphous silicon (α-Si) technology, the MO based TFTs (MOTFTs) are reported to have much higher electron field-effect mobility (μFE) due to the large, spherical ns-orbitals (n≥4). Moreover, the MO semiconductor materials also have their advantages in transparent applications due to the wide bandgap (~3 eV). Therefore, the wide-bandgap MO semiconductors including indium (In), gallium (Ga), zinc (Zn), and Tin (Sn) based binary or multi-component oxides have gradually become the promising channel material candidates for advanced TFT based technologies. However, for the well-established vacuum-based MOs fabrication technologies such as magnetron sputtering, atomic layer deposition (ALD), and chemical vapor deposition (CVD), etc., the complex processes, high-demand equipment, and small depositing area heavily limit the development for low-cost MOs deposition. Therefore, the solution process, one feasible and facile route to deposit the MO films under an ambient condition was proposed and reported by the researchers. Nevertheless, every coin has two sides, there is often a trade-off between the low cost of solution methods and the high performance of TFTs. The solvent residues or incomplete annealing process may lead to the defects and ruin the performance of the devices. Accompanied by the challenges, many studies have been reported to deduce the side effects brought by solution process and plenty of breakthroughs also have been done. In another word, to fabricate the high electrical performance TFTs based on low-cost solution processes still have great room for development and are worthy of study. In this work, for environmental protection and cost reduction considerations, we mainly focus on the spin-coating based n-type In-free semiconductor zinc tin oxide (ZnSnO, ZTO). We firstly proposed a kind of deionized (DI) water solvent-based fabrication routine for ZTO semiconductor films. The fabrication process was operated under a low temperature (≤ 300℃) in air condition. Combining with the silicon dioxide (SiO2) dielectric layer, the TFTs with a μFE of 2 cm2V-1s-1 were successfully fabricated. Furthermore, with the help of the novel two-dimensional (2D) material MXene, we tuned the work function (WF) of the ZTO channel and optimized both the μFE (13.06 cm2V-1s-1) and the gate bias (GB) stability behaviors of the TFTs via depositing the homojunction structured channels. Subsequently, we replaced the SiO2 dielectric with the solution-processed high-k aluminum oxide (AlOx) films, the devices showed an increased μFE of 28.35 cm2V-1s-1 and applied to a resistor-load inverter successfully (Chapter 2). Secondly, besides the performance optimization, the solution-processed TFTs could also be applied to realize the advanced high-parallel neuromorphic network computing tasks. The TFTs that could meet this application requirement are regarded as the synaptic transistors (STs) and are decided to mimic the biological synapse. The operating basis is established on the hysteresis window in STs’ transfer characteristics and non-volatile multi-level variable channel conductance. Here we applied the MXene to the interface between the ZTO channel and the SiO2 dielectric layer and proposed a kind of floating-gate transistors (FGTs) with the functions of STs. The MXene induced FGTs (MXFGTs) successfully mimicked the typical behaviors of biological synapse under both the gate voltage (VGS) and channel incident ultraviolet (UV) light stimuli. To further explore the suitability of the MXFGTs in machine learning task, we utilized the classifier based on the artificial neural network (ANN) and the tested results of the devices to simulate the image classification process. The training and recognition results of the images based on the Modified National Institute of Standards and Technology (MNIST) database further proved the application potential of MXFGT in neural network (NN) system (Chapter 3). Finally, in Chapter 4, we further improved the light detecting behavior of the MXene based STs. A shell layer of germanium oxide (GeOx) was grown to cover the nanosheets of MXene through a facile solution method. The obtained GeOx-coated MXene (GMX) nanosheets were doped into the ZTO channel layer and fabricated into the GMX based STs (GMXSTs). Owning to the area enlarging function of the high electron density MXene core and the heterostructure of GeOx/ZTO bilayer, the GMXSTs showed excellent optoelectrical synaptic performance under visible light stimuli, which was highly improved over MXFGTs. Then, we applied the various responses of the devices under the different input lights into image target area detecting simulations. With the help of the detecting pre-process, the tasks of counting the fluorescent cells stained by 2-(4-Amidinophenyl)-6-indolecarbamidine dihydrochloride (DAPI) was correctly performed. Finally, the "night vision"-inspired and the brightness-adjusted image reconstruction results were presented, which further indicated the bright future of this kind of synaptic device in the application field of artificial visual perception.

Abstract:The objective of this project is to investigate the use of information-visualisation interfaces to support co-located synchronous collaboration in large display multi-device environments. This is interface for people working together, at the same time in the same place, using interactive graphics with multiple connected computer devices, including large-displays as well as handheld devices such as smartphones and tablets. The technology to support this type of working is already commonplace and research indicates there are a significant advantages to this mode of working. There are however some significant challenges for the designers and developers of these types of system. Firstly, these interfaces need to deal with the inherent complexity of human-human as well as human-computer interaction. We have a limited understanding of how users work together in multi device environments, and need to manage a shared display space and shared control to accommodate different types of collaborative working. For example, at different times users can be working together closely, working separately, or even in disagreement. This introduces an additional level of complicity to the design of the interface that makes it in-practical to simply adapt traditional visualisation interfaces or apply established visual interface design methodologies for multi device collaboration. The work in this thesis investigates the processes of co-located synchronous collaboration in large display multi-device environments and how we can improve this type of activity through better interface design. par The first stage in our study, to improve the design of large display multi-device environment interfaces for co-located collaboration, was to observe how people tend to work in this type of environment using a simple interface with basic view coordination on a large display connected to two mobile devices. This revealed that the users would tend to move between patterns of individual and closer collaborative working. The next stage of our study was to look at ways to better support this type of working by making the transition between individual and collaborative working smoother for the user. This included looking at different ways of coordinating views between mobile devices and the large display, managing different user selections, and showing different users. Finally we used the results of these studies to develop an interface for a real-world collaborative visualisation activity. This validated some of the design decisions generated and demonstrates the potential of large display multi-device collaborative visualisation. Our study demonstrated the potential of a large-display to act as a shared space for users to view the context of their selections and share their insights into the data. We also present preliminary design guidelines on how to coordinate selections among the different displays for different types of collaborative activities. This includes the finding that users preferred a single shared view on a large display for close collaboration or activities that lend themselves naturally to turn taking and a more predictable split screen with separate views for activities that involve a large navigable area and more independent working. Users also preferred to have selections automatically sent to the large screen and interact to retrieve the selections of other users. It was also found that weak presentation of non-essential information was the most effective method for highlighting different user selections. These findings can help guide interface designers and developers for this sort of environment in the future.