16168 research outputs found
Sort by
Towards optical flow in time-of-flight range imaging using the continuous wavelet transform
No full textTime-of-flight (ToF) range imaging cameras perform full-field distance measurement, but require multiple frames to be captured sequentially in order to construct a single range image. When motion occurs during data acquisition, artifacts appear that limit the accuracy of range estimates. Analysis of these artifacts finds that they take the form of spatial waves whose frequency encodes motion information. We therefore propose a new technique for measuring motion in Time-of-Flight range imaging based on the continuous wavelet transform (CWT). Synchrosqueezing is explored as a means of improving the accuracy of velocity estimates
The safetyfication of curriculum as an example of bureaucratic governance
No full textThis presentation takes up what Author 1 terms the “safetyfication” of education. Safetyfication refers to the process through which the notion of ‘safety’ has acquired an omnipresent status in education in the Anglophone world as a result of an amalgam of claims that range from disciplinary calls to social justice demands. In this paper-presentation, the authors will analyze attempts to make the curriculum safer – or free from sociopolitical risk – under Author 2’s political analysis of bureaucracy as a form of governance (2018; 2023). The main argument made is that this form of governance operates inadvertently, through the safetyfication of education
Child past and future thinking: What children remember and imagine, and the role of parental mind-mindedness and elaboration
No full textThinking about memories and possible future events allows us to prepare for the future. Whilst we are beginning to understand how past and future thinking develops in early childhood, little is known about these processes in middle childhood and how parental factors, like mind-mindedness and elaborations, influence them. This study examined the content of past and future thoughts in 11- to 13-year-olds and the role of these parental factors. We collected data from 21 parent-child dyads in three different phases: a) parents completed an online survey involving a measure of mind-mindedness; b) parents and children completed a joint Zoom session measuring parental elaborations; c) children completed an independent Zoom session measuring past and future thoughts. Children generated more relevant details for past thoughts than future thoughts, while the amount of off-topic details was similar between past and future thoughts. This pattern indicates that children are proficient at retrieving relevant information, but they are more proficient at this retrieval when thinking about memories than imagining the future. Children subjectively rated past and future thoughts as similarly important and vivid but reported thinking more frequently about future thoughts than past thoughts. This pattern illustrates that, for children, memories and future events are equally important and visualised equally. However, children imagine the future more often than they reminisce about the past. These findings enhance our understanding of the functioning of past and future thinking in middle childhood. No relationship was found between parents’ understanding of their children, parents’ elaborations, and children’s past and future thoughts. We replicated and extended previous findings with a novel age group, thereby contributing to a lifespan overview of past and future thinking. Our research has practical implications for parents, teachers, and clinicians. A longitudinal design in future research would further assist in understanding past and future thinking across childhood
Exploring the antecedents and consequences of corporate environment performance for sustainable development
No full textCorporations are significant contributors to climate change, and increasing public awareness has led scholars to investigate Corporate Environmental Performance (CEP) and its relationship with sustainability. The empirical research on CEP remains limited, particularly regarding the antecedents (such as Board Composition Index) and consequences (such as Corporate Financial Performance) of CEP with impacts of several moderating factors (such as stakeholders, social media sentiment, corporate reputation, and national culture). This study aims to fill the gaps in extant literature by employing a multi-country dataset and rigorous econometric analysis to examine the objective of this thesis – exploring the antecedents and consequences of CEP. Corporate Environmental Performance, Board Composition Index and Corporate Financial Performance (CFP) are the key variables that define the main hypothetical relationships explored in this thesis. CEP refers to corporate efforts to mitigate environmental risks, optimise resource utilisation, and reduce emissions. Board Composition Index refers to a composite of attributes of a corporate board, and Corporate Financial Performance refers to the measurement of a company's financial health and its ability to generate profits. In addition, this thesis investigates the moderating impact of several factors, such as community and regulatory stakeholders, on the relationship between BCI and CEP; the moderating impact of culture, management control strategies (MCS), Corporate Reputation and Social Media Sentiment on the relationship between CEP and CFP. The research is structured into three key parts, representing three articles that align with the key focus of the thesis. The first article investigates the antecedents of CEP, focusing on the composition of the board of directors and stakeholders' interests in CEP. Stakeholder theory is the primary framework used in this article. The second article explores the consequences of CEP by analysing social media sentiment and corporate reputation. Drawing on the Resource-Based view, the article illustrates how firms can achieve a competitive advantage through enhanced CEP in dynamic environments. In addition, Attribution Theory is employed to explain public perceptions of corporate sustainability efforts and their influence on social media sentiment, which in turn affects environmental performance. Article three examines the roles of national culture and proactive management control strategies in shaping CEP outcomes. All the articles in the thesis use panel datasets employing different regression techniques to test the proposed models. The thesis provides both theoretical and practical contributions. The research contributes to extant literature by offering valuable theoretical and practical insights that would interest academics and practitioners. Through a comprehensive methodological approach, including developing a Board Composition Index (BCI) and examining the joint impacts of stakeholders, the study provides new insights into how these variables influence CEP. Moreover, exploring the joint effect of social media sentiment and corporate reputation brings new understandings of how firms’ environmental performance influences financial performance. Additionally, exploring cultural influences and environmental management strategies deepens the knowledge of their role in the CEP-CFP relationship, offering practical guidance for managers in multinational corporations
Evaluating and using remote sensing data for studying economic activity and population change in China
No full textuse of remote sensing data, especially night-time lights (NTL) data, has flourished
in recent decades as a proxy for measuring economic activity and the spatial distribution
of population. The availability of these data can make up for the absence of and possible
inaccuracy in traditional economic statistics, especially because NTL data are available
at fine scale and with high frequency. The growing use of these data raises the question
of how accurate is this night-time lights proxy, and for what purposes is it best suited
when studying economic development and population changes at sub-national levels.
In the first part of this thesis, a 20-year time-series of GDP at China’s third
sub-national level (counties, county-level cities and districts) and census population
counts at the same level from 2000, 2010 and 2020, are used as a benchmark to examine
the performance of multiple sources of NTL data as proxies for local economic activity
and inequality. I also test the accuracy of gridded population estimates, some of which
rely on NTL data, for estimating inter-census changes in local population.
Based on the comprehensive evaluation of different NTL sources at different levels of
spatial aggregation, the second part of the thesis uses NTL and other remote sensing
data, in conjunction with the traditional GDP and population statistics, to study a set of
economic issues that are pertinent to regional development in China. The impact of
administrative upgrading of county-level units (a type of regional policy practiced in
China) and of modern transport infrastructure is studied, using both GDP and NTL data
to measure economic impacts. These studies use spatial econometric models that allow
for spillovers. The changes in the city size distribution, in terms of land and people, is
also studied because of the possibility that China’s cities are physically expanding in
places other than where the population is moving.
Overall, the thesis contributes a nuanced understanding of how NTL data and
other remote sensing data can assist in studying complex social and economic problems,
as well as raising some caveats about the limits to the uses of such data
Photography within te ao Māori: Decolonising the photographic lens
No full textŪkaipō: Wānanga, places of origin and photographic representation.
Ūkaipō recognises the cultural, emotional, and spiritual along with the physical ties a Māori has with their place of origin, even if it is not their place of birth, emphasising grounded forms of connection between people and place and subsequent understandings of places and spaces. Hence the importance today of people asserting their pēpeha, tohu whenua landmarks with which they most strongly identify, consolidating their ‘birth of place’ or ‘place of birth’
Monitoring te mana o te wai: Intergrating an indigenous cultural health index with conventional biomonitoring tools for improved freshwater management
No full textGlobally, there are universal efforts to quantify and address human impacts on ecosystems. In particular, changes in land use and river regulation have led to drastic declines in stream and river ecosystem health. These human impacts are driving the global freshwater biodiversity crisis and degrade ecosystem services provided by aquatic habitats. In Aotearoa New Zealand there is widespread concern about the state of our Nation’s freshwater ecosystems. Coupled with a renaissance in Māori culture, there has been a drive to develop indicators that help articulate cultural values, assess ecosystem health from a cultural perspective, and provide greater agency for Māori in environmental monitoring and management. In my thesis, I sought to integrate an established cultural monitoring framework with conventional biomonitoring tools for improved freshwater management. I hypothesized that using a suite of monitoring approaches would help detect the impacts of human land use on stream ecosystem health, and that cultural indicators would be consistent with conventional measures in diagnosing land use impacts. I also hypothesized that there could be discrepancies between cultural monitoring and conventional approaches, as each represent different perspectives and ways of knowing streams and rivers.
To test my hypotheses, I selected ten stream sites across the catchment of the Kuratau River, a major tributary of Lake Taupō. Based on River Environment Classification land cover types, I used a balanced study design with five sites draining indigenous forest, and five sites draining pastoral land. I measured key stream physicochemical properties and collected water samples for determination of nutrient concentration and faecal coliform counts. I also characterised instream and riparian habitat. I collected benthic macroinvertebrate samples and sampled water for environmental DNA (eDNA) to assess different facets of biodiversity. I measured decomposition using the Cotton Strip Assay and quantified periphyton biomass using a portable fluorometer. In combination with these approaches, I also used a Cultural Health Index, an established cultural monitoring framework that assesses three components: customary significance, taonga species/mahinga kai (food gathering), and the Cultural Stream Health Measure (CSHM) to assess identified attributes for stream health.
I found that pastoral land uses lead to a decline in stream ecosystem health when compared to sites draining indigenous forest. Pastoral sites had higher specific conductivity and concentrations of total nitrogen and nitrate. Habitat variables changed, with increased sedimentation and more degraded riparian zones in pastoral streams. Decomposition rates as measured by the cotton strip assay increased in the pasture sites, which was likely a response to increased nutrient availability. There were shifts in macroinvertebrate composition, with key indicator taxa for Indigenous forest sites including pollution-sensitive mayflies and stoneflies. Total taxa, EPT (Ephemeroptera, Plecoptera, Trichoptera), and aquatic insect richness all significantly declined in pastoral sites. Increases in primary production were less pronounced, but did increase alongside abundances of grazing caddisflies and snails in the sites draining pastoral catchments.
The Cultural Health Index indicated that sites in the Kuratau River catchment had customary and mahinga kai values. CSHM responded negatively to pastoral land uses, and the effect size it described was congruent with other responses. The CSHM was strongly correlated with an indicator of riparian habitat condition (the Riparian Condition Index) and had weak negative correlations with sedimentation and stream temperatures. The CSHM was not correlated with macroinvertebrate indices, which may have been because of the relatively few sites sampled, the narrow impact gradient in the Kuratau catchment, and the lack of temporal replication. However, cultural monitoring approaches should be seen to complement conventional environmental monitoring methods for assessing freshwater and not be expected to duplicate them. The strength of the Cultural Health Index approach comes from establishing Māori agency to assess the environment through a Te Ao Māori lens, and to support and foster inclusion for tangata whenua epistemologies in environmental monitoring and resource management. Future research should consider greater hapū involvement and further explore the mechanisms driving changes in stream health and biodiversity
The influence of income on subjective well-being: An empirical study in transitional country of Vietnam
No full textAbstracts
The pursuit of happiness is not merely a personal endeavour but also a societal imperative, influencing governmental policies and individual well-being. In the realm of economics, the correlation between income and happiness has long been a subject of scrutiny, particularly the phenomenon known as the income paradox. This paradox, first highlighted by Easterlin in 1974, questions why despite rising incomes, average happiness levels often remain stagnant. Numerous studies have since delved into this puzzle, with a predominant focus on relative income and income inequality. However, there remains a notable research gap on this subject in transitional countries, especially Vietnam.
This thesis aims to enrich the empirical study of the influence of income on subjective well-being, in part to elucidate the income - happiness paradox in Vietnam. This thesis includes three papers.
The first paper employs panel data sourced from the Viet Nam Access to Resources Household Survey (VARHS), spanning three periods — 2012, 2014, and 2016. Utilising fixed effect estimates, it assessed the impact of absolute income and three specifications of external comparison income on individual happiness. The focal point of this paper lies in categorizing subgroups based on relative income rather than absolute income. This approach enables the removal of conflicting relationships, allowing for a clearer understanding of how social comparison income influences individual happiness. Findings suggest that individuals compare themselves to multiple reference groups, both upwards to those better off and downwards to those worse off. However, statistically significant results emerge only when one comparison outweighs the other, or when both comparisons align either upward or downward impact. Intriguingly, the outcome shifts from asymmetry to near symmetry when the interactive effects among reference groups change from opposition to alignment.
The second paper presents a pioneering investigation into the impact of self-perceived economic status on subjective well-being. This study employs panel data sourced from VARHS, covering three periods: 2016, 2018, and 2020. By employing fixed effect estimates, it investigates the dynamic interplay between absolute and relative self-perceived household economic status within the commune and subjective well-being. This exploration is conducted by applying four primary specifications and an analysis of the interaction between increases in per capita income and shifts in household economic status. The findings reveal that present self-perceived economics status positively influences personal happiness, while past self-perceived economics status has the opposite effect.
Additionally, the influence of comparisons between present and past self-perceived relative income can exhibit symmetry or asymmetry, depending on the approach. Moreover, a positive relationship is identified between the distance in a household's current and past economic household rank and individual happiness. External comparison is asymmetrical and is a stronger factor than internal income comparison. Despite an increase in absolute income, overall personal SWB is not guaranteed to rise, underscoring the pivotal role of the perception of household economic position or relative income in determining individual happiness.
The third and final paper represents one of the earliest studies to examine the influence of self-perceived income inequality on life satisfaction, employing an instrumental variable approach. It is also a pioneering study in utilising data from the VARHS of 2020 in investigating the effects of self-perceived income inequality on life satisfaction in Vietnam. The results highlight the adverse impact of perceived income inequality on individuals' overall life satisfaction, with controls for endogeneity further emphasizing this effect. These findings underscore the importance of government initiatives aimed at reducing economic disparities to enhance individual subjective well-being
Optimal decision trees via search: A reinforcement learning framework
No full textDecision Trees are one of the most popular models in Machine Learning because of their interpretability, which is especially noticeable for Decision Trees with few splits (or decision rules), a property that is called sparsity. Interpretable models are highly valued in domains where decisions carry substantial ramifications such as health-care and the criminal justice system. For this reason, seeking optimal sparse Decision Trees constitutes a fundamental problem in Interpretable Machine Learning, and a large research effort has been devoted to solving this problem.
Due to the NP-Hard difficulty of the problem, greedy heuristic methods such as C4.5 and CART have been favoured historically, and while fast and scalable, these methods remain unsatisfactory because they forgo the sparsity portion of the problem. In fact, they often lead to suboptimal and overly complex Decision Trees. Nevertheless, many algorithms have been developed in the literature to address the sparsity problem, with recent breakthroughs leveraging Dynamic Programming (DP) and Branch & Bound (B&B) techniques, and causing a paradigm shift from traditional Mathematical Programming to search algorithms. However, most of the proposed DP and B&B methods follow a Depth-First-Search (DFS) strategy, which, while attractive due to its storage-economy capacity, is inefficient because of its uninformative nature. Moreover, DFS necessitates the definition of a maximum depth hyperparameter a priori, and its inefficiency becomes a limiting factor when considering large maximum depths. These concerns were partially addressed with Best-First-Search (BFS) strategies. BFS is appealing because it prunes the search space more aggressively without losing the optimality guarantee, thus enabling it to find the optimal solution faster than DFS. Moreover, BFS does not necessitate fixing a maximum depth hyperparameter, it can run on infinite depth, which is an important advantage that allows practitioners to focus tuning efforts on other hyperparameters. However, these advantages come at the expense of higher memory consumption than DFS. To alleviate this issue, the search strategies of BFS methods have to be formulated as efficiently as possible, in other words, they need to find the optimal solution as quickly as possible. The reason being to tip the balance of their advantages-drawbacks more towards their advantages, outweighing their high memory consumption issue.
The current BFS methods can be improved substantially by reconsidering the sparsity problem under a framework that benefits from its structural properties. Our main contribution in this thesis addresses this concern within a Reinforcement Learning framework. Our BFS algorithm, Branches, benefits from an AO*-type search strategy, which, coupled with our pruning bound we call Purification Bound and several other heuristic considerations, focuses the search on relevant regions of the search space while maintaining the optimality guarantee. This enables Branches to find optimal sparse Decision Trees faster than the state of the art, and as a result alleviate the memory burden that weighs on BFS approaches. We analyse the computational complexity of Branches theoretically by deriving an upper bound on the number of branch evaluations it performs before terminating, and we show this result to be superior to similar analyses from the literature. These findings are further validated through a series of extensive experiments showing Branches to significantly advance the state of the art in terms of number of iterations, runtime, anytime behaviour and scalability potential. This work lays the foundations for future work that can further improve the scalability of these methods.
In addition to this work, we also study the online setting where a data stream is observed rather than a fixed dataset. Unlike the batch setting, the literature on optimal sparse Decision Trees in is scarce for Online Learning. Most approaches adapt greedy heuristics to handling data streams through statistical estimates, hence inheriting their suboptimality issues. On this front, we develop three Monte Carlo Tree Search (MCTS) algorithms with asymptotic optimality guarantees, and we show that they can successfully retrieve the optimal solution in situations where the state of the art methods are practically guaranteed to fail. Furthermore, we adapt Branches to the online setting inducing Online-Branches, an algorithm that satisfies finite-time optimality guarantees rather than asymptotic ones. Our experiments show Online-Branches to outperform our previously developed MCTS methods. Our work on the online setting has the potential to draw more research attention to the problem of seeking optimal sparse Decision Trees when given a data stream. It also lays a solid foundation for these future works through several discussions about the potential issues that arise due to the hypotheses of Online Learning and our proposed solutions
The athleticism of drag queen practices and performance: A mixed-methods study of injury, prevention, and health care
No full textMost research on drag performance has focused on gender and sexuality and/or performing arts.
To date, very little has examined drag queens’ experiences of athleticism, injury, and injury care. This research provides insights into the physical experiences of professional drag queens, and the relationship of drag queens to health care, in Aotearoa New Zealand and the United States. Specific areas of inquiry include: how drag queen performers of diverse gender identities, ages, experience levels, and athletic backgrounds experience the use of their bodies in their practice of drag; injury occurrence related to drag queen performance; and how drag queens perceive, access, and experience health care in direct connection to their drag practice.
A mixed-methods approach was employed, combining a survey (96 participants) and interviews with six drag queens from Aotearoa New Zealand the U.S. This research draws on Queer and feminist theory, as well as public health models, to inform data collection and analysis techniques. The findings reveal drag to be a physically demanding discipline: the majority of drag queens have experienced drag-related injuries, despite practicing personally developed body maintenance behaviours. Participants also expressed a lack of, and necessity for, widely available, fully accessible, and visible LGBTQ+ and drag culturally safe and resonant healthcare. These insights contribute to a broader understanding of the physical body in drag queen performance, as well as the health care needs of drag queens in Aotearoa New Zealand and the United States. Utilization of research findings, along with cross-sector discussions, could inform community health best practices and policy development, to improve health care for this historically marginalized population