Search

Inorganic salts are widely used medium- and high-temperature phase change materials (PCM). However, their thermal conductivity is low, and since salts are highly corrosive, many conventional heat transfer enhancement methods do not apply. In this research, the ceramic foam was used to enhance the heat transfer of salts. The ceramic foam is open-cell, and its pore size is large (~mm), which enables the natural convection of PCM and benefits thermal transport. And it is fabricated alone and is integrated with molten salt easily; hence, the foam and salt can be maintained separately (rather than the whole CPCM), and the maintenance cost is reduced. This research covers property measurement, material preparation, compatibility test, and performance analysis. The thermo-physical properties of nitrate salts with various component fractions were measured. Mixing nitrate salts can decrease the melting point without impairing the latent heat; a molecular dynamics simulation was performed to provide microscopic insights. The ceramic foam/salt composite phase change material (CPCM) with excellent compatibility was fabricated; the thermal transport in the porous CPCM was studied experimentally and numerically. The melting front is significantly curved, indicating natural convection is strong, which is attributed to the low viscosity of salts. The melting rate of the CPCM with a pore density of 10 PPI (Pores Per Inch) is increased by 51.5%, and the porosity has a more significant effect than pore density on the melting performance. For the shell-and-tube heat storage unit, the ceramic foam with varying outer diameters is optimal in melting rate, total stored energy, and energy storage rate. The thermal performance under solar fluctuation is also studied; strong solar radiation not only improves the energy storage rate but also increases the total stored energy; compared to pure PCM, ceramic foam/salt CPCM is more sensitive to solar fluctuation.

Information management systems are being developed and introduced to engineering design activities to help companies reuse their information. This trend has been identified in multi-location companies, which operate design departments at various locations. Investigating how multi-location companies can configure their information management environment to fulfil engineers' information needs in design process opens up a research topic for us. A well configured information management environment will require a clear understanding of what designers need from it. A literature review has identified that only a few studies have observed the way in which design practitioners work in an industrial environment; and since these studies were all conducted 10 or 20 years ago within a specific context, their findings may no longer reflect current practice. The first phase of this research investigates the information needs of designers in the vehicle industry, and the different approaches and resources that they currently adopt to fulfil these needs. The findings of this first phase of research are then used to identify further industry-based studies. The second phase of the research focuses on the structures and operations of information management for engineering design. A case study of four multi-location companies was conducted to understand their information management environments (IMEs). Primary data was collected in fieldwork visits to the companies. Inductive grounded coding was applied to analyse the data, revealing the constructs of information management environments at the case companies studied. The analyses identified strategic orientations, structures, organisational enablers and individuals' capabilities in case companies' IMEs. A new classification of IMEs was proposed to understand their structural and operational features. The analyses have implications for how companies should configure their IMEs. The studies reported in this dissertation contribute to theory by providing an understanding of the configuration of IMEs in companies and proposing ways of configuring IMEs in practice. Researchers in the domain of information systems can develop a good understanding of how professional practitioners interact with information environments so that they can propose information management systems or methods that make tangible improvements. This study also helps engineers map out the information environment that they search to acquire information. Chief engineers or managers in companies who are in charge of information management can benefit from the understanding of their own information environment and use the proposed model as a guide to configure their own information management environments. The study also suggests future research directions, such as identifying and proposing the indicators that can be used to measure the performance of information management.

Talent Management (TM) is still a new research field in China’s context. TM like other management disciplines is often embedded in a certain institutional context and influenced by certain social norms, cultural factors and government policies. TM is therefore closely related to research context. The Chinese cultural context is influential in shaping TM concepts and practices. However, how TM practices are used to retain talents in the context of Chinese private-owned enterprises (POEs) is still a research gap. TM is a holistic strategy for an organisation. It is therefore necessary to explore Chinese POEs’ TM practices through a holistic lens, which covers the processes of defining talents, attracting talents, developing talents and retaining talents. Previous TM studies mainly focused on exploring TM concepts and there is a lack of empirical investigations on TM practices, especially in the context of Chinese POEs. Linked to the characteristics of Chinese POEs and Chinese cultural context, this study explores talent retention from a holistic perspective of TM. This research adopted an interpretivist perspective and inductive approach. Based on a case study research method, primary and secondary data were collected from three case companies, and analysed qualitatively. The theoretical framework used in this research is largely based on TM literature. Talent attraction, talent development, and talent retention are used as a lens to view the case companies and to explore what TM activities drive talent retention in the context of Chinese POEs. Organisational commitment is the main theory employed by this research to explore talented individuals’ turnover intention. Findings show that Chinese guanxi is an important perspective to define talented employees in the context of Chinese POEs. Competence, position, and guanxi are holistically considered to define a talented employee. It was found that guanxi is an important factor influencing the entire TM process, including attracting talents, developing talents and retaining talents. Career development, rewards, and guanxi were significant factors in retaining talented employees in the context of Chinese POEs. Guanxi as a new TM perspective not only makes a theoretical contribution to talent definition but also contributes to talent development and talent retention theories. The research offers practical talent retention suggestions to TM practitioners. A significant practical contribution may be adopting TM practices to develop talented employees’ guanxi ties to increase their organisational commitment and to reduce turnover intention.

AMP-activated protein kinase (AMPK) plays important roles in the regulation of energy homeostasis and has been implicated in a number of diseases including diabetes and cancer. Obtaining a full picture of AMPK signalling is important for the understanding of these disease states and the development of novel therapies. AMPK is activated by phosphorylation of T-172 within the α catalytic subunit. While a number of AMPK upstream kinases have been identified, identification of the AMPK phosphatase has proved elusive. In the first part of this thesis, the identity of the AMPK phosphatase(s) was investigated. The data presented shows that treatment of mammalian cells with the protein phosphatase (PPP) family inhibitor Calyculin-A caused marked increases in AMPK activity. These results indicate that one or more of the AMPK phosphatases belong to the PPP family, with the concentration profile pointing towards PP1 involvement. However, using siRNA based techniques; it was not possible to pinpoint the specific phosphatase(s) that dephosphorylates AMPK. Our results, combined with recently published studies, suggest that regulation of AMPK dephosphorylation is a complex process with involvement of different protein phosphatases depending on cell-type and/or conditions of cell stress. Next, quantitative mass spectrometry techniques were utilised in order to identify AMPK interacting proteins. From the list of candidate interactors, the interaction between AMPK and the PAK1 signalling complex was characterised. AMPK directly phosphorylated PAK1 on Ser-21 leading to its activation. Activation of AMPK in mammalian cells led to increased activity and phosphorylation of endogenous PAK1. PAK1 activity/phosphorylation was reduced in cortical neurons derived from AMPK α1 knockout mice compared to wild-type neurons. Furthermore, activation of AMPK with A769662 in melanoma cell-lines significantly increased cell invasion and this effect was abolished by the PAK1 inhibitor, IPA3. These results show that PAK1 is an AMPK substrate and suggest that at least part of the effects of AMPK on axonogenesis and tumour formation is mediated through PAK1 signalling. In addition to the effects of AMPK on PAK1, PAK1 was shown to directly phosphorylate the AMPK α1 subunit at Thr-377 and Ser-403/405. This phosphorylation was able to further increase the activity of CaMKKβ phosphorylated AMPK. This finding adds a novel layer of control to the regulation of AMPK.

Information cooperation, reuse and integration can be developed on the platform of rapidly growing open distributed environments and can support development of Ambient Intelligence. However, in such environments, information may be only partially observed due to the unreliability of data collection technologies and heterogeneity in the ontologies employed caused by distributed and independent system development. These challenges need to be overcome to facilitate intelligent data analysis. We focus on the use of large-scale databases such as statistical databases and data warehouses, where aggregates can be obtained to summarise information; such aggregates are valuable in providing efficient access, computation and communication. A principle-based learning framework is proposed and developed for semantically heterogeneous aggregate data using maximum likelihood techniques via the EM (Expectation-Maximisation) algorithm. The learning framework inherently handles data incompleteness and schema heterogeneity from unreliable, incomplete or uncertain information sources. The framework is developed for supervised and unsupervised learning from data in a distributed environment. This development is demonstrated using two scenarios. In the first scenario a decision-making mechanism is proposed to support assistive living for elderly people in a smart home environment. The mechanism incorporates modules for learning inhabitants' activities of daily living based on partially observed and unlabelled data, enabling hierarchical activity prediction and assisting inhabitants in completing activities by providing personalised reminders. Real data have been collected in a smart kitchen laboratory, and realistic synthetic data are also used for evaluation. Results show consistent and robust performance and other information and insights are also obtained. In the second scenano a model-based clustering algorithm is proposed for independently developed distributed heterogeneous databases to support cooperation between organisations, including distributed smart homes from different institutions. Clustering in the presence of data heterogeneity enables the characteristics of similar contexts to be captured. The algorithm is systematically evaluated using simulated data, with encouraging results and good scalability to large numbers of databases.

High-frequency acoustic waves (above MHz) are widely applied in ultrasonic imaging and many microfluidic applications. For many cases, the response of interfaces to high-frequency acoustic waves can be simplified as a reflection and refraction process, where only the effect of the interfaces on acoustic wave propagation is considered. But for some physical processes, such as bubble and droplet manipulation, the response of the interfaces plays a key role. However, the interaction processes between high-frequency acoustic waves and interfaces, especially fluid interfaces, are mostly highly nonlinear. Most of the related work focused on acoustic waves with frequencies ranging from DC to kHz. There are continuously new phenomena found and reported showing the effect of high-frequency acoustic waves on fluid interfaces contrary to previously discovered theorems and laws. Yet many of the phenomena are poorly understood and explained. We briefly introduced the background of high-frequency ultrasonic devices, including the selection of material, design principles, physical models, and some of the important applications. Observations and measurements on the interface dynamics with small amplitudes and high frequencies have always been a challenge to researchers. Gas vesicles (GV) are bubbles wrapped by a protein layer with submicron sizes, which put forward an even higher requirement for the equipment and experiment design. We reported a precise non-contact technique to monitor GV's vibrational behavior using laser Doppler vibrometry. The resonance frequencies of the single GV and agglomerated GV were found and verified by the simulation results. We then focused on the behavior of the GV affected by lower frequencies (1-10 MHz) associated with medical and industrial applications. The change of GV's vibrational response to the acoustic wave with varying power input was observed. The mechanism of GV's function on ultrasonic signal amplification can be well explained with the measured linear and nonlinear behaviors (buckling and collapse). We then investigated the effect of high-frequency acoustic waves on the fluid-gas interface on the droplet's surface. We found that the sessile droplets with a small volume ( ~ μL) produce visible, low-frequency vibrations (~ 100Hz)when exposed to high-frequency acoustic waves. However, there are no appropriate theories to predict capillary wave generation excited by high-frequency acoustic waves since the frequencies violate fundamental assumptions used in these theories. We brought up the acoustic radiation pressure-interface shape feedback mechanism to explain the energy transfer across vastly separate scales. The simulations based on the physical model are carefully compared with the direct observation of droplet interface dynamics from the high-speed digital holographic camera. The pressure-interface feedback model accurately predicts the vibration amplitude threshold at which capillary waves appear, the subsequent amplitude and frequency of the capillary waves, and the distribution of the standing wave pressure field within the sessile droplet responsible for the capillary waves.

Recommender system (RS) is a useful information filtering tool for guiding users in a personalized way of discovering products/services from a large space of possible options. A good recommender system can not only ameliorate the prevalent issue of over-choice due to information explosion but can also promote sales and boost revenues. As such, it has become a vital and indispensable component in the modern internet industry. In this thesis, we aim to tackle several key prevalent tasks in recommender systems, including (1) modeling the complex relationships between users and items; (2) integrating side information; (3) modeling the temporal dynamics; and (4) cold-start problem. To this end, we investigate neural architectures and geometric methods, aiming to tackle those challenges and improve recommendation performances. Recently, deep learning's revolutionary advances in speech recognition, computer vision, and natural language processing have gained significant attention. Many challenging problems such as feature representation learning and sequence modeling can be well addressed with deep neural models. Our key contributions on personalized recommendations with neural architectures are listed as follows: Firstly, we investigated feature representation learning for hybrid recommender systems, which involves learning items' feature representation with a contractive autoencoder. We proposed AutoSVD and AutoSVD++ to efficiently anticipate user's ratings for unseen items. Secondly, we investigated the personalized ranking task and proposed NeuRec and its extension, SNR, to model the complex and nonlinear relationships between users and items. Thirdly, we investigate the challenging cold-start problem and present a model that considers both item and user side information under a unified deep neural framework which can introduce personalization and collaborative effects to cold-start recommendations. Fourthly, we proposed a self-attentive metric learning framework to tackle the temporal dynamics in recommender systems. Beyond neural networks, we also investigate geometric methods which concern modeling the intricate relations between users and items with geometric inductive biases. We proposed FML to factorize the explicit Euclidean distance with standard learning algorithms. We also proposed a collaborative filtering model using Quaternion representations. Those methods achieve comparable performances as neural networks with simpler formulation.

IoT environments are increasingly being used to support critical infrastructure, placing an increased burden of responsibility on these systems to perform reliably and safely. IoT environments, however, are often laden with security and reliability issues relating to sensor performance and network security. This Thesis presents research towards engineering safer and more reliable IoT environments. Within this Thesis, a review of the state-of-the-art in reliability is performed to guide the solutions investigated. Part one of the Thesis concerns the area of sensor reliability. A study is performed to measure the impact of sensor failures in a smart environment. This study concluded with recommendations for sensor data preparation and model selection. Following this, a novel approach is presented for detecting anomalies in constrained sensors. This approach achieved an f-measure of 92.00% on a public activity recognition dataset. Part two of the Thesis considers the network layer of the IoT, and in particular the area of intrusion detection. A study is performed to develop a hierarchical intrusion detection system, which successfully lowers the number of false alarms recorded by the system to 0.59%. Next, multi-task learning is investigated for intrusion detection systems, and a model is proposed using balanced batch learning. This model achieves a macro average recall of 92.94%. This Thesis clearly demonstrated that solutions to improving reliability in the IoT are possible, at both a sensor and network level.

It is expected that by 2050 153 million people globally, predominantly elderly, will be living with dementia. As the proportion of older people increases, we need to find innovative ways to support their care. With the easy availability of inexpensive wifi-enabled cameras and the recent development of open-source tools to generate accurate pose-estimation models using video data, automated pose-based activity recognition becomes possible, even for sedentary behaviours. Early research focused on generating features from 3D models built using data from specialised RGB cameras embodying additional depth sensors. However, conventional RGB cameras are an attractive option for data capture due to their low cost and ease of use. We show that 2D models can be as informative as 3D models as a basis for sedentary activity recognition, thus opening the way for the development of affordable and accessible systems to support care at home for people with dementia. This thesis explores the potential of using skeletal keypoints in body and hand models developed from video data to automatically identify agitation in people living with dementia, assessing the frequency and intensity of agitated episodes. As many older adults are largely sedentary, we focus on agitated behaviours displayed whilst a person is seated within view of a single RGB video camera. Agitation can manifest itself in several different ways, including repetitive mannerisms such as scratching, picking, or rubbing. Using hand and wrist keypoints from pose estimation models generated from video, we learn to recognise repetitive hand movements using a small number of discriminative poses. Our method, which is the first to use hand keypoints for recognising non-communicative gestures, is invariant to the speed of hand movement, recognising nearly 80% of the types of hand movement. Moreover, when some user-specific information was included in the training data, our model identified over 98% of hand movements. In addition to learning to recognise agitated behaviour, recognising changes in the intensity of agitation can provide important metrics for monitoring changes in behaviour. By first learning characteristics of the movements corresponding to particular agitated behaviours, our model was able to detect over 90% of changes in the intensity of repetitive hand movements. Although activity recognition based on pose estimation models has attracted much interest, most work has focused on recognising activities from pre-segmented video sequences. However, identifying and assessing agitation requires the detection of (often occasional) abnormal behaviour from within long periods of continuous data. Correspondingly, we also present a method to segment streamed data automatically into sequences containing similar types of hand movements. Our approach improves the detection of agitation by 67% relative to the moving window approach used by other agitation detection models. Additionally, an effective method of cleaning hand keypoint data is proposed. The findings of this research will support the development of unintrusive systems for monitoring agitated behaviour. Automatic monitoring will enable better understanding of agitated behaviours in dementia, providing accurate and objective metrics for clinicians to determine appropriate care pathways, and enable carers to be alerted when intervention is required.

A non-convex sparsity promoting penalty function, the transformed L1 (TL1), is studied in optimization problems with its applications in compressed sensing (CS) and matrix completion. The TL1 penalty interpolates l_0 and l_1 norms through a nonnegative parameter a \in (0,+\infty), similar to l_p with p \in (0,1]. TL1 is known in the statistics literature to enjoy three desired properties: unbiasedness, sparsity, and Lipschitz continuity.For compressed sensing problems, a RIP condition for TL1 exact recovery is proposed and proved. Next, difference of convex algorithms for TL1 (DCATL1) are presented in computing TL1-regularized constrained and unconstrained problems. For the unconstrained problem, we prove convergence of DCALT1 to a stationary point satisfying the first order optimality condition. In numerical experiments, we identify the optimal value a=1 and compare DCATL1 with other CS algorithms. An explicit fixed point representation is also developed for the TL1 regularized minimization problem. The TL1 thresholding functions are in closed form for all parameter values. The TL1 threshold values differ in subcritical (supercritical) parameter regime where the TL1 threshold functions are continuous (discontinuous) similar to soft-thresholding (half-thresholding) functions. We propose TL1 iterative thresholding algorithms and compare them with hard and half thresholding algorithms in CS test problems. The TS1 penalty, as a matrix quasi-norm defined on its singular values, interpolates the rank and the nuclear norm through a nonnegative parameter a \in (0, +\infty). We consider the unconstrained TS1 regularized low-rank matrix recovery problem and develop a fixed point representation for its global minimizer. The TS1 thresholding functions are in closed analytical form for all parameter values. We propose TS1 iterative thresholding algorithms and compare them with some state-of-the-art algorithms on matrix completion test problems.

Methionine (Met) is an essential amino acid for animals and also the first limiting amino acid in a broiler diet. The dietary supplemental Met sources include the natural isoform L-methionine (L-Met), the synthetic form DL-methionine (DL-Met) and the synthetic Met precursor DL-2-hydroxy-4 (methylthio)-butanoic acid (DL-HMTBA). The objective of this dissertation was to determine the effect of different dietary Met source supplementation and Met deficiency on a series of physiological and biochemical aspects, including growth performance, global DNA methylation and methyltransferase activity, blood antioxidant profile (e.g., acute phase protein, leukocyte count), intestinal nutrient transporter gene expression, Met converting enzyme gene expression and activity, oxidative stress markers, and a potential pathway related to amino acid signaling. To achieve this goal, male Cobb-500 broilers were raised from day of hatch (d0) to d35 post-hatch and fed a diet deficient in methionine and cysteine (Met + Cys) (control) or the same diet supplemented with 0.22% DL-Met, 0.22% L-Met or 0.31% DL-HMTBA (to provide 0.22% DL-Met equivalent) to meet Met + Cys requirements. Tissues (liver, breast muscle, duodenum, jejunum and ileum) and blood samples were collected at various ages from d0 to d35 for analysis. Met supplementation significantly improved body weight gain and feed efficiency compared to the Met deficient group, but no differences were observed among DL-Met, L-Met and DL-HMTBA for growth performance parameters (P > 0.05). Met supplementation had no effect on red blood cell packed cell volume, white blood cell differential count, hepatic total DNA methylation, DNA methyltransferase and Met oxidase activity, and had limited effects on activation of p70S6K, a key amino acid signaling protein (P > 0.05). Although dietary Met sources did not change oxidative status of the treated chickens, both L-Met and DL-Met but not DL-HMTBA supplementation decreased the level of acute phase protein serum amyloid A compared to the control group (P > 0.05). The effect of Met supplementation on gene expression of nutrient transporters and Met converting enzymes were complex and dynamic. Most of the target genes demonstrated tissue- and development-dependent expression patterns, with few significant treatment effects observed. L-Met and DL-Met but not DL-HMTBA supplementation enhanced the neutral amino acid transporters ATB0,+ and B0AT gene expression in various small intestinal segments. All three Met sources increased monocarboxylic acid transporter (MCT1) gene expression in the jejunum. DL-HMTBA and L-Met fed chickens showed greater hepatic L-HMTBA oxidase (HAO1) gene expression. DL-Met increased glutamic-oxaloacetic transaminase 2 (GOT2) gene expression in the duodenum. An in vitro study with tissue explants, however, did not demonstrate a similar gene expression pattern as that in the in vivo study. Lastly, RNA sequencing results illustrated that Met deficiency could lead to many differentially expressed genes but different supplemental Met sources had no influence on hepatic gene expression profiles. In conclusion, as common dietary supplemental Met sources, L-Met, DL-Met and DL-HMTBA exhibited similarity in impacting intestinal amino acid/peptide/monocarboxylic acid transporter gene expression and Met converting enzyme activity. The regulatory roles of Met as an antioxidant and nutrient signaling in cell metabolism were not affected by different dietary supplemental Met sources.

In the deep interiors of planets, the pressures and the temperatures are significantly higher than at ambient condition. The interplay between compression and thermal expansion leads to complexity in the chemistry and physical properties of minerals. Improved knowledge of the behavior of dominant minerals inside the Earth and other planets is essential for understanding the structure, dynamics, and evolution of planetary systems.In this dissertation, I study materials in environments ranging from the deep mantle of the Earth to the core boundary of Jovian planets and to stellar interiors, using computer simulation with first-principles methods. The contents of this dissertation are as follows: Chapters 1-2 provide an overview of Earth and planetary minerals, and a background of first-principles computer simulation methodology. Chapter 3 discusses the thermoelasticity of iron- and aluminum-bearing MgSiO$_3$ in the Earth's lower-mantle. Chapter 4 is about novel phases of hydrogen-oxygen compounds at giant-planet interior conditions. Chapter 5 presents an equation of state of warm dense sodium. We conclude that lower-mantle mineralogy is in accord with the pyrolite model; we demonstrate that minerals can have counter-intuitive stoichiometry and phases at high pressure; and we provide a coherent first-principles equation of state table of sodium in a wide range of density and temperature conditions. This work illustrates the importance of first-principles simulations as powerful tools in Earth and planetary science studies. Finally, chapter 6 discusses possible directions for future work in the field of computational mineral physics.

Due to recent advancements and the incessant progression of wireless sensor networks, conducting Human Activity Recognition (HAR) research within smart environments has become a widely explored domain. Nevertheless, whilst extensive research has been carried out, HAR remains a highly intricate and challenging task. Each stage of the data-driven HAR process contributes to the overall performance, thus, optimisation within each stage has driven research endeavors. This Thesis presents an end-to-end methodology for the optimisation of HAR, which involves investigations into enhancing performance at various key stages of the process. A publicly available HAR dataset was utilised throughout to evaluate and demonstrate the effectiveness of the proposed approach. Initial explorations focused upon the pre-processing stage, within which the impact of data quality upon activity classification was explored using data-driven approaches to HAR. Findings demonstrated the negative impact of noise upon classification performance, with a significant performance increase of 12.97% when using cleaned data. This work led to providing recommendations as to how data should be pre-processed to prevent reductions in performance. Subsequent explorations focused upon enhancing HAR performance during the feature selection stage, within which a new hybrid feature selection method was produced. Findings revealed the effectiveness of the developed method which achieved an enhanced HAR performance of 83.24%, in addition to demonstrating the benefits of performing feature selection. A considerable trade-off was revealed between the classification performances achieved and the number of redundant features identified and removed, in comparison to the evaluated well-stablished feature selection techniques. Finally, research endeavours focused upon optimising HAR performance during the classification stage, within which both novel homogeneous and heterogeneous ensemble methods were produced. Findings demonstrated the effectiveness of the proposed ensembles, in particular the heterogeneous method which outperformed 4 benchmarked classifiers achieving an overall classification performance of 84.13%.

Activity recognition is a research area focused on identifying human actions from a collection of sensor data. Often, solutions rely on design-time techniques that remain static after being trained once and assume that further data derives from identical sources and conditions. However, real-world applications require solutions to embrace daily living's unpredictability and allow for exposure to unexpected changes. Transfer learning is a growing field that can address the variability of activity recognition problems by reusing the knowledge from previous experiences to recognise activities from different conditions, resulting in the leveraging of resources such as training and labelling efforts. While integrating ubiquitous sensing technology and transfer learning seem promising, there are some research opportunities that, if addressed, could accelerate the development of activity recognition. This Thesis reviews the current status and challenges of off-the-shell capabilities for transfer learning opportunities in modelling activities of daily living in classification problems. It starts by introducing a novel method consisting of fusing features at the hidden level of artificial neural networks derived from different activation functions to create a more informative set of features. This method's finding shows that the fusion of two activation functions contributes to scenarios with small quantities of labelled data available in the source domain and the absence of labelled data in the targeted domain. It then builds upon the learned experience from the previous method to investigates the benefit of feature fusion as a strategy to enhance the Learner model within the Teacher/Learner approach when the feature-space is different between domains. The Thesis concludes by presenting a framework that converges the feature fusion strategy with the Teacher/Learner approach over the active learning technique to automatise the selftraining process of the Learner models. Evaluation of these studies shows promising results towards the mitigation of the insufficiency of labelled data available by enabling the Learner model to outperform the Teacher's performance, as presented in the Chapters within this Thesis.

The ability to adapt fuel usage to nutrient availability is termed metabolic flexibility, and is influenced by activity of the pyruvate dehydrogenase complex (PDC). The Virginia lines of chickens are a unique model of anorexia and obesity that have resulted from 56 generations of artificial selection for high (HWS) or low (LWS) juvenile body weight. We hypothesized that hyperphagia and obesity in juvenile HWS chickens are associated with altered fatty acid oxidation efficiency and metabolic flexibility in tissues associated with energy sensing and storage, and relative cellular hypertrophy in white adipose tissue. Hypothalamus, liver, Pectoralis major, gastrocnemius, abdominal fat, clavicular fat and subcutaneous fat were collected from juvenile (56-65 day-old) HWS and LWS chickens for metabolic, gene expression and histological assays. The HWS chickens had reduced fatty acid oxidation efficiency in abdominal fat (P < 0.0001) and reduced rates of oxidation in abdominal fat and gastrocnemius (P < 0.0001) as compared to LWS. There was reduced citrate synthase activity in white adipose tissue (P < 0.0001) and greater metabolic inflexibility in skeletal muscle (P = 0.006) of HWS compared to LWS. Greater pyruvate dehydrogenase kinase 4 (PDK4) and forkhead box O1 (FoxO1) mRNA were found in skeletal muscle and white adipose tissue of 56-day-old HWS than LWS. Expression of peroxisome proliferator-activated receptor γ (PPARγ) in all adipose tissue depots was greater (P < 0.05) in LWS than in HWS chickens. The HWS chickens had larger (P < 0.0001) and fewer (P < 0.0001) adipocytes per unit area than LWS. These results suggest that the HWS chickens have impaired metabolic flexibility and fatty acid oxidation efficiency due to an up-regulation of pyruvate dehydrogenase activity to accommodate the influx of acetyl CoA from fatty acid oxidation in skeletal muscle and white adipose tissue. These metabolic adaptations can be linked to differences in gene expression regulation and body composition between the lines. Adipocyte cellularity data are consistent with greater oxidative efficiency in the adipose tissue of LWS, because of the greater number of unfilled cells in all depots that were sampled. Results can be extrapolated to agricultural production in the understanding of factors regulating the amount of lipid deposition in chicken carcass fat. Results may also provide insight into eating disorders and the development of obesity.

Abstract: The goal of error control coding is to encode information in such a way, that in the event that errors occur during transmission over a noisy communication channel or during storage in an unreliable memory, the receiver can correct the errors and recover the original transmitted information. Low-density parity-check (LDPC) codes are a kind of high performance linear block code, which are already used in many recent communication systems. Information rates guaranteed by these codes approach the theoretical Shannon capacity limit. In addition, LDPC codes are now widely used in practice and included in many communication standards. Therefore study of these codes is very important and practical approaches to their design and decoding techniques are of great interest. There exist very good decoding algorithms for LDPC codes with respect to different channel models. However, LDPC codes suffer from the infamous error floor problem at high signal-to-noise ratios. This problem is attributed to the trapping sets by Richardson. It is shown that the dominant trapping sets of regular LDPC codes, so called absorption sets, undergo a two-phase dynamic behavior in the iterative message-passing decoding algorithm. We present a linear dynamic model for the iteration behavior of these sets. It can be seen that they undergo an initial geometric growth phase which stabilizes in a final bit-flipping behavior where the algorithm reaches a fixed point. Our analysis is shown to lead to very accurate numerical calculations of the error floor bit error rates down to error rates that are inaccessible by simulation. The topologies of the dominant absorption sets of two example codes, the IEEE 802.3an [2048,1723] regular (6,32) LDPC code and the Tanner [155,64,20] regular (3,5) LDPC code, are identified and tabulated by using topological features in combination with search algorithms, respectively. To make our analysis more complete, we provide more solid evidence showing that trapping sets are equivalent to absorption sets. In other words, we argue that absorption sets characterize all failure mechanisms. Some insights from our linear analysis can be borrowed to approach this problem. Another insight that this formula provides to us is the means to reduce the error floor, which is another important part of what we hoped to achieve by developing the formula in the first place. By allowing the log-likelihood ratios (LLRs) utilized by the message-passing decoding to grow bigger in precision length, the absorption sets as a trouble-causing structure will be successfully corrected. Therefore, the absorption sets, generically born with the code design, will no longer threaten the error correcting performance as long as the messages have enough room and time to grow. This translates to greater LLR clipping thresholds and more iterations, both of which are preset at the decoder. However, the actual settings are dependent on how much the subgraph of the dominant absorption set resembles that of a non-zero minimum-weight codeword, therefore the settings are empirical. In general more likeness in their topologies, the more effort it costs to correct the absorption set.

Activity Monitoring is a key feature of health and well-being assessment that has received immense consideration from the research community over the last few decades. In recent years, smart phones with inbuilt sensors have become popular for the purpose of activity recognition. The sensors capture a large amount of data, which contain meaningful events, in a short period of time. Hence, the capability to detect, adapt and respond to such changes performs a key role in various domains such as to identify changes in patient vital signs in a medical domain or to assist in the process of generating activity labels for the purposes of annotating real-world datasets. The sudden change in mean, variance or both may represent a change point in time series data. A change point can also be used to identify the transition from one activity to another. Change point detection is a technique to process and analyse the sensor data and identify the transition from one underlying time series generation model to another. In this thesis, the existing Multivariate Exponentially Weighted Moving Average (MEWMA) algorithm has been used to automatically detect such change points for transitions in user activity. The MEWMA approach has the advantage that it does not require any assumptions to be made in relation to the underlying distributions to evaluate multivariate data streams and can run in an online scenario. Following this, the genetic algorithm (GA) has been used to identify the optimal set of parameters for a MEWMA approach to change point detection. The GA optimizes different parameters of the MEWMA in an effort to find the maximum F-measure, which subsequently identifies the exact location of the change point xiv from an existing activity to a new one. Furthermore, we benchmark our approach against a similar multivariate approach, namely Multivariate Cumulative SUM (MCUSUM) to automatically detect change points in different user activities. In addition, GA and Particle Swarm Optimization (PSO) are also used to automatically identify an optimal parameter set using different parameters for MEWMA and MCUSUM, so as to maximize the objective function that is Fmeasure. The evaluation is performed using different metric measures based on real and synthetic datasets collected from accelerometer sensor. The experimental results shows that the proposed approach MEWMA outperforms than the bench mark approach MCUSUM. Hence, the accurate change point detection in the data enable a system to identify changes in user activities and recognize and monitor good behaviour such as healthy exercise patterns based on these activities.