Search

This thesis explores the relationship between the corporate governance and earnings management in director's background and the relationship between the CEO and the controlling shareholder. Moreover, as this study is based on the Chinese market, and there some difference from west market. Therefore, this study will also discuss the difference Chinese market and west market, and improve the Jones model to measure earnings management in Chinese market. This study tend to address the following questions. First, how to measure earnings management for Chinese listed companies using Chinese accounting standard? Second, whether the directors' background influences the firm's earnings management. Third, whether the relationship between CEO and controlling shareholders has an impact on the firm's earnings management, all empirical chapters focus on Chinese listed companies. Firstly, previous research about earnings management uses the Jones model and various Jones-based model (like the modified Jones model, performance-adjusted model) to measure company's earnings management. These models were developed based on Western accounting principles, which is different from the Chinese accounting standards. There is some measurement error when using these models to measure the company's earnings management in the Chinese market. This study develop a new model based on Chinese new accounting standards, which was issued in 2007. The results indicate that the new model developed is better than previous models used in literature. Secondly, based on the new model developed about how to measure earnings management for Chinese listed companies, this study investigate the impact of the director's financial background on the company's earnings management. The results show that the effect on the company's accrual earnings management is negative. In contrast, the impact on the company's real activist's earnings management is positive when the director has some financial education or work experience. These results indicate that directors with financial background tend to help the controlling shareholder to manipulate earnings through real activities rather than accruals. Finally, this study investigate the impact of the relationship between the CEO and controlling shareholder on the company's earnings management. Same as the above, this study use the new model measuring earnings management. The results show that when the controlling shareholder's shareholding at a low level, there is a conflict between the CEO and controlling shareholder. However, as the increasing of the controlling shareholder's shareholding, the CEO tends to collaborate with the controlling shareholder to manipulate the company's earnings. This study developed a measurement model more suitable for the characteristics of the Chinese market, making it more accurate and effective. Based on the data analysis of the annual reports of domestic listed companies, the degree and scale of financial manipulation can be detected more effectively. This study suggests that policymakers pay particular attention to the background of directors. A financial background gives directors the ability to help companies manage their earnings. Policymakers also need to pay attention to the relationship between the CEO and the controlling shareholders. The CEO tends to help the company to do earnings management if he/she is working for the controlling shareholder.

This thesis analyses the intersection of migrant workers and non-government organisations (NGOs) in China and India through a relational-spatial framework. The high political, economic and social significance of migrant workers and NGOs in both countries make it essential to have more detailed insights into how the two communities relate to each other, but literature to-date tends to give priority to macro-structural visions, and consequently leave micro-level realities and the role of multiple actors largely untouched. I therefore integrate human geography approaches from Henri Lefebvre and Doreen Massey into a new spatial framework, and then conceptualise NGO-migrant interaction through this framework to interpret NGO-migrant relations from their outset. Based on 18 months of in-depth qualitative fieldwork on eight NGOs across four cities in China and India, I explore the contextualisation of NGO-migrant relations in each country, and highlight their similarities and differences. In both countries I find that there is a political-economic regime which strives to preserve its dominance over workplaces and public places, to which NGOs in both countries are subject; however, the authoritarian regime in China is more restrictive of NGOs' attempt to construct interactive space than the more liberal-democratic one in India. As a result, Chinese NGOs' strategies and outcomes of their relational formation tend to be much more polarised than those of Indian NGOs, with different types of NGOs in different cities showing different responses. This thesis highlights the significance of contextual impacts on NGO-migrant relation, and showcase the usefulness of a relational-spatial framework in exploring nuanced details within multi-actor dynamics, with important implications for research on labour, voluntary sector, and development in countries with different political systems.

Reconstructing the surface microtopography through data fusion from a range of microscopies offers a way to provide considerably more information than any one technique alone. This thesis explores several methods for combining the info rmation from different kinds of microscopes including height data, electron images and optical images. Each microscopy has distinct advantages and disadvantages that can be considered in terms of data leakage in the transfer function from the actual sampl e topography to any individual dataset. These leakage paths are identified for a range of microscopies including atomic force microscopy, scanning electron microscopy and white light interferometry. Existing techniques for reconstruction from multiple image datasets, such as shape-from-shading and stereomicroscopy are evaluated, again in terms of information leakage. A new approach is presented which attempts to minimize leakage by identifying a pathway of data fusion which first isolates the most reliable information in each dataset, and then maintains this information as the datasets are combined. The identification of the most trusted data requires knowledge of each imaging mechanism, while the final reconstruction is based on iterative matching between the datasets and a simulated image. Hence the final reconstruction incorporates both the datasets from the microscopes and their imaging physics to suppress data leakage. This approach is first validated with a model micromachined sample using a combination of height interferometer data together with back-scattered and secondary scanning electron microscope images. It is then applied to particulate samples to demonstrate the capabilities of the approach.

With the expeditious advancement of information technologies, health-related data presented unprecedented potentials for medical and health discoveries but at the same time significant challenges for machine learning techniques both in terms of size and complexity. Those challenges include: the structured data with various storage formats and value types caused by heterogeneous data sources; the uncertainty widely existing in every aspect of medical diagnosis and treatments; the high dimensionality of the feature space; the longitudinal medical records data with irregular intervals between adjacent observations; the richness of relations existing among objects with similar genetic factors, location or socio-demographic background. This thesis aims to develop advanced Statistical Relational Learning approaches in order to effectively exploit such health-related data and facilitate the discoveries in medical research. It presents the work on cost-sensitive statistical relational learning for mining structured imbalanced data, the first continuous-time probabilistic logic model for predicting sequential events from longitudinal structured data as well as hybrid probabilistic relational models for learning from heterogeneous structured data. It also demonstrates the outstanding performance of these proposed models as well as other state of the art machine learning models when applied to medical research problems and other real-world large-scale systems, reveals the great potential of statistical relational learning for exploring the structured health-related data to facilitate medical research. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]

This project is the first in-depth study to investigate the pluralistic soundscapes of the Guerx Sal Lad, one of the most important religious festivals of the Bai in the Dali Bai Autonomous Prefecture, Yunnan, southwest China. “Guerx Sal Lad” is the name of the festival in the Bai language, which means “visiting three places.” Held annually (from the twenty-second day to the twenty-fifth day of the fourth lunar month), the Guerx Sal Lad encompasses a variety of practices—the worship of tutelary deities, pilgrimage, and antiphonal singing, among others—that are related to people’s religious beliefs, wishes for prosperity, and maintenance of intracommunity relations. As a festive space, the Guerx Sal Lad presents a desired situation by the state regarding China’s religious policy, in which folk beliefs are accessible and even celebrated as cultural heritage but, at the same time, with the strong regulatory presence of the state ideology and policies implemented by local governments. It is within the Guerx Sal Lad that these different interests intersect. Based on ethnographic fieldwork undertaken in Dali city from 2020 to 2022 and analysis of documentary sources including government documents and scholarly writings, I theorize that the soundscape of the Guerx Sal Lad is composed of a polyphony of overlapping and intersecting sub-soundscapes that represent the voices and ideologies of several key actors involved in shaping the festival: rural villagers, local government institutions, and mediators in between (basic-level cadres, representative transmitters, and local scholars). By looking into various sounding events closely related to the Guerx Sal Lad and discourses around them, this dissertation explores the negotiation and contestation between these “voices” with unequal “volumes” at the intersection of minority folk beliefs, China’s minority politics, and state projects of cultural heritagization, especially since China’s engagement with UNESCO’s Convention for the Safeguarding of the Intangible Cultural Heritage (ICH) in the early 2000s. Informed by ethnomusicology, sound studies, religious studies, and the large body of scholarly research on ethnic minority groups in the People’s Republic of China (PRC), this dissertation offers a new approach to understanding minority politics and religious practices in contemporary southwest China.

The scale of machine learning problems grows rapidly in recent years and calls for efficient methods. In this dissertation, we propose simple and efficient methods for various large-scale learning problems. We start with a standard supervised learning problem of solving quadratic regression. In Chapter 2, we show that by utilizing the quadratic structure and a novel gradient estimation algorithm, we can solve sparse quadratic regression with sub-quadratic time complexity and near-optimal sample complexity. We then move to online learning problems. In Chapter 3, we identify a weak assumption and theoretically prove that the standard UCB algorithm efficiently learns from inconsistent human preferences with nearly optimal regret; in Chapter 4 we propose an approximate maximum inner product search data structure for adaptive queries and present two efficient algorithms that achieve sublinear time complexity for linear bandits, which is especially desirable for extremely large and slowly changing action sets. In Chapter 5, we study how to efficiently use privileged features with deep learning models. We present an efficient learning algorithm to exploit privileged features that are not available during testing time. We conduct comprehensive empirical evaluations and present rigorous analysis for linear models to build theoretical insights. It provides a general algorithmic paradigm that can be integrated with many other machine learning methods.

Abstract: This thesis utilizes a combined numerical and machine learning approach to explorethe performance of an alumina ceramic tile undergoing high-velocity impact. The fi-nite element model is established by incorporating a user-defined Johnson-Holmquist-Beissel (JHB) material model within the framework of smoothed particle hydrody-namics (SPH) in LS-DYNA finite element software. The computational frameworkis validated across a range of conditions by matching the simulation results fromboth plate impact experiments and ballistic testing from the literature. Once vali-dated, the model is used to generate training data sets for an artificial neural network(ANN) to predict the residual velocity and projectile erosion of an alumina ceramictile undergoing high-velocity impact in the SPH framework. The ANN is then used toperform a sensitivity analysis involving exploring the effect of mechanical properties(e.g., strength and shear modulus) and impact simulation geometries (e.g., the thick-ness of ceramic tile) on material performance (i.e., residual projectile velocity anderosion). Overall, this study shows the capability of the hybrid FEM-ANN approachin studying the high-velocity impact on ceramic tiles and is applicable to guide thestructural-scale design of ceramic-based protection systems.

Due to the continuous aging process in China, not-for-profit residential care institutions (NPRIs) account for about half of the service providers in aged care services and are playing an increasingly significant role. However, several NPRIs have been exposed to governance scandals, such as the corruption of senior management; NPRIs using their not-for-profit status as an empty shell to generate profit for illegal distribution; NPRI managers’ unlawful use of government subsidies; and extreme situations such as fraud, crime, and aged abuse cases, which has contributed to the public accountability requirements in this sector. This research, therefore, asks: Can Chinese NPRIs’ existing corporate governance systems fulfill their accountability function? If not, how can they be enhanced to achieve greater accountability to protect the interests of stakeholders? To address the research question, this research developed an accountability framework by analysing the NPRIs should be accountable to whom; accountable for what; and how to improve the accountability. This accountability framework is used in analysing the related legislation and governance documents with four key accountability mechanisms in NPRIs, including the monitoring role of the supervisory board; the evaluation of the related party transactions; the transparency of the annual report; and the public enforcement has taken by regulators. This research identified several potential system flaws and proposes some specific recommendations to enhance the accountability mechanisms. For example, the legislator should value the accountability demands of clients; the legislator should emphasis the formal and functional independence of the supervisory board ; the Regulatory regime should consider regulating related party transactions with due process and directors’ loyalty duty rules; the NPRI practitioner should think of improving the transparency of annual reports to outsiders with higher levels of understandability and accessibility; and regulators should improve the current public enforcement system with more enforcement tools and a greater use of them to respond to the violations appropriately. By implementing the proposed improvements, this research hopes that the current corporate governance system could be more accountable to broader stakeholders of NPRIs, in both substantive and procedural issues, especially to protect the most important and vulnerable stakeholders, aged residents.

This dissertation comprises three independent studies to examine the causes and implications of CEO and executive turnover. Chapter 1 argues that Knightian uncertainty, frequently conceptualized as “unknown unknowns,” is not only an exclusive attribute of a firm’s external environment but also an attribute of any managerial decision (e.g., executive turnover). This study employs minimax regret equilibrium to present a mathematical model of individual decision-making under conditions of uncertainty when the probabilities of future states are not objectively known. We propose an alternative account of senior executives’ preferences under conditions of uncertainty, discussing the escalation of commitment, stigma avoidance, herding in the voluntary exit, and risk propensity; these factors have often been characterized as irrational, as they systematically violate the axioms of expected utility theory. Chapter 2 challenges the presumption of agency theory that CEOs are likely to entrench themselves unless they are forced out or raided. Building on this assumption, substantial studies have emphasized the board of directors’ effectiveness in assessing leadership efficacy and removing incompetent or self-serving CEOs to eliminate agency costs. I theorize that CEO career concerns result in anticipated board interventions, which potentially contribute to agency costs. Theoretically, unnecessary board interventions may prompt more CEO departures, which disrupt organizational routines and impose managerial talent loss. I empirically examine the value-destroying implications of CEO turnover in relation to the elevated likelihood of board interventions at the firm level. This paper addresses a form of agency costs that result from CEO career concerns and CEO-board power dynamics.Chapter 3 examines executives’ preferences for external job opportunities as opposed to staying conditional on relative firm performance and relative changes in firm performance. Previous tests of this theory have emphasized corporate governance’s role in involuntary executive departure and executives’ diminished post-departure career outcomes. The results of this study highlight the conditions under which executives are attracted to external job opportunities that are not randomly assigned to them. The findings complement our understanding of managerial labor market prospects after voluntary turnover.

Teams have arguably been the most essential organisational form in human society. During the pursuit of better team performance, several factors have attracted consistent attentions; among which, trust has been widely recognised as especially important. Trust directly impacts team performance as it plays a crucial and pivotal role in the decisions that each team member (each agent) makes with regard to its own actions and its interactions with fellow team members (other agents). However, there is no systematic and generic modelling of trust aiming at improving team performance by enabling each agent in the team to conduct trust-based interactions through accurate trust evaluation and prescribing the appropriate actions in response to that trust evaluation. This thesis addresses this absence of such a framework in three steps. Firstly, objective interaction records (past one-to-one interactions between pairs of agents) are considered the most reliable source of information pertaining to trustworthiness. Thus, to obtain accurate trust evaluations in a team of agents, a Determination of Trust Model (DoTM) is proposed. This DoTM models trust by establishing the relationship between interaction data and the trust values through a machine learner. Secondly, a Novel Trust Architecture (NoTA) is proposed to address the appropriate transition from trust to practical actions for an agent conducting an interactive team task. A team task may involve multiple sub-tasks and different types of interactions, so determining the appropriate interactions according to trust is challenging as it requires the differentiation of the trust (of the agent being interacted with's capacity) in performing different sub-tasks and the association between trust and designated interactions. The NoTA is a nuanced framework determining an agent's appropriate actions during the task through the application of differentiated trust and matched response strategies. Finally, an integrated trust model is proposed on the basis of the DoTM and the NoTA, which enables an agent the full autonomy of trust-based interactions. Experimental demonstration is conducted in two selected domains, i.e. a food foraging task (FFT) and a coverage task. In each domain, agents are differentiated as reliable agents and flawed agents in terms of their trustworthiness; with the assumption that each scenario only involve one type of flawed agent, experiments are conducted in scenarios possessing agents with different types of flaws and ratios between flawed and reliable agents within a team. The DoTM shows an average evaluation accuracy (ACC) of at least 94% in the FFT domain, and 80% for 14 types of flawed agents (out of 15) in the coverage task. It is also shown that with a priori knowledge about trust, by applying the NoTA, average team performance is improved by at least 5% (and up to 331%) for 7 types of flawed agents (out of 12) in the FFT domain compared with the baseline team performance. In the coverage task, the improvement is at least 17% (with a maximum of 431%) for 7 types of flawed agents (out of 15). By applying the integrated trust model, when agents possess no a priori knowledge about each other's trustworthiness, it is shown that in the FFT domain, with regard to 5 of the types of flaws (out of 12), average team performance is improved by at least 6%. In the coverage domain, in presence of 7 types of flaws (out of 15), team performance is improved by at least 13%. Based on experimental investigations, it can be concluded that (1) the proposed DoTM enables high trust evaluation accuracy in a team with agents possessing different trustworthiness; (2) the proposed NoTA is capable of obtaining the optimal strategy which determines the appropriate actions during interactive teamwork according to trust for optimised team performance; (3) without a priori knowledge about trust, the proposed integrated trust model facilitates improvement to team performance in various scenarios with agents having different trustworthiness. The proposed computational trust models can be applied to or used as an important tool to improve the performance of a team of intelligent agents conducting a variety of cooperative/collaborative tasks.

During the past decades, femtosecond laser micro-fabrication has gained growing interests owing to its several unique features including direct and maskless fabrication, flexible choice of materials and geometries, and truly three-dimensional fabrication. Moreover, fiber-optic sensors have demonstrated distinct advantages over traditional electrical sensors such as the immunity to electromagnetic interference, miniature footprint, robust performance, and high sensitivity. Therefore, the marriage between femtosecond laser micro-fabrication and optical fibers have enabled and will continue to offer vast opportunities to create novel structures for sensing applications. This dissertation focuses on design, fabrication and characterization of optical-fiber based interferometric devices for sensing applications. Three novel devices have been proposed and realized, including point-damage-based Fiber Bragg gratings in single-crystal sapphire fibers, all-sapphire fiber-tip Fabry-Pérot cavity, and in-fiber Whispering-Gallery mode resonator

The primary objectives of this study are to explore the brand personality of birdwatching, and to investigate how sound interventions alter participants’ selfimage/birding brand personality (SI/BP) congruities and participants’ self-image/birders’ image (SI/BI) congruities, as well as their intention to participate in birding and introduce birding to others. A quasi experiment was designed for college students within a mid-western university. Results of the present study show that the brand personality of birding is patient, nature lover, outdoorsy, and discovery; whereas the perceived image of birders is independent, intelligent, nature lover, and outdoorsy. The study also indicates that the congruence levels between how a person looks at himself/herself and the perceived images of birding have a significant influence on his/her intention to participate in birding and introduce birding to others. The sound intervention has a significant influence on participants’ intention to participate in birding and introduce birding to others. The results from this study could be helpful to both tourism marketers and birding club leaders.

Decreased strength of trabecular bone is a direct effect of osteoporosis, which can be evaluated by finite element analysis. However, computational limitations have restricted previous trabecular bone analyses primarily to the linear domain. In addition, previous work was largely invasive and the corresponding finite element models were typically homogeneous. Nonlinear heterogeneous finite element analysis was used to calculate trabecular bone apparent strength directly from in vivo micro computational tomography (micro-CT) scans. Through a series of validation experiments, it was shown that this nonlinear modeling is more accurate in evaluating trabecular bone mechanics in osteoporosis than previous work. A parameter driven set of material properties was employed in the finite element models using gray levels in the form of Hounsfield units as the independent variable. This enabled the finite element models to capture the variations of material properties of trabecular bone. The methods and techniques for converting the micro-CT scans into finite element models, defining the finite element models (element type, material properties characteristics) and solving the models are discussed in detail. In addition, the techniques developed for image preprocessing, such as image registration and image degradation, are also provided in this dissertation. The scanning, image processing and modeling methods and techniques were applied to two groups of rabbits, an ovariectomy group and a control group, to evaluate the time-course of trabecular bone osteoporosis. Our experiment showed that ovariectomy significantly slows the normal bone strength increase over time observed in the control group. The strength increase over time was due to a combination of increased bone architecture indices such as volume fraction and trabecular thickness as well as increased material properties due to greater bone tissue density. Compared to heterogeneous models, the homogeneous models reflected less strength increase over time because they lack the capability to capture tissue level material property variations. Volume fraction analysis alone resulted in even lower predicted increases in bone strength because it could only monitor the bone apparent level density variation. Thus the nonlinear heterogeneous models, with parameter driven material definitions, are more accurate than other types of models or methods.

Task planning is a key component necessary to implement telerobotic systems in remote operations for the deactivation and decommissioning (D&D) of defunct nuclear facilities. In this thesis, a task planning software module for a telerobotic D&D system is presented. This software helps the operator to visually create a plan for a dismantlement task so that the control component can then interpret the task plan file and execute the plan. The graphical user interface allows the operator to plan a set of tasks efficiently. This software also calculates the manipulator's end-effector orientation and position at approach points, cut points and final points through appropriate kinematics transformations and saves the data to a task plan file. With this software, the D&D telerobotic system works as an integrated system with real-time multi-mode operation. It is believed that this task planning concept will be applicable to other application domains.