39456 research outputs found
Sort by
Beyond Recognition: Indigenous Sovereignty and Equity in Ontario's K-12 Landscape
No full textWhile efforts have been made to include culturally relevant, anti-oppressive, and decolonizing pedagogy and curriculum within Ontario’s public education system, these efforts take place within a system built to reproduce its own colonial values. Grounded in Néhiyaw (Cree) principles of research and informed by broader theories of resurgence and refusal, this thesis explores conceptual tensions between equity and Indigenous sovereignty in Ontario’s K–12 educational landscape. Rejecting the colonial politics of recognition, it uses an Indigenous Literature Re-view methodology with four interrelated phases: Searching, Analysis, Yarning, and Re-view. This research traces policy narratives that shape educational discourse in Ontario. What emerges is a vision of equity that is highly reliant on a theory of change based on the politics of inclusion and is limited within frameworks prioritizing student achievement. This framing confines Indigeneity, neutralizing resurgent possibilities and rendering settler colonialism and land invisible
Effective Black Holes in Quantum Gravity
No full textThis dissertation explores quantum gravity applications to resolve black hole singularities, investigates the potential transition from black holes to white holes, and characterize the resulting spacetime in each scenario. The following work utilizes various approaches, including canonical loop quantum gravity (LQG) and the generalized uncertainty principle (GUP), while also adopting a more flexible framework that is not tied to any single quantum gravity theory. Instead, this research relies on guiding principles expected to emerge from a complete theory of quantum gravity. Through this exploration, the dissertation aims to contribute to the resolution of key issues in black hole physics within a quantum gravity setting.
The first significant contribution of this thesis involves the application of the Raychaudhuri equation to assess the existence of singularities. In particular, loop quantum gravity corrections to the Raychaudhuri equation are derived for the interior of a Schwarzschild black hole and near the classical singularity. The analysis demonstrates that the resulting effective equation leads to the defocusing of geodesics, caused by the emergence of repulsive terms. This effect prevents the formation of conjugate points, renders the singularity theorems inapplicable, and ultimately facilitates the resolution of the singularity in this spacetime.
Building on this, the next part of the work extends similar ideas within the framework of the generalized uncertainty principle. To address challenges identified in a previous model of the interior of a generalized uncertainty-inspired black hole, an "improved scheme'' is introduced, drawing inspiration from loop quantum gravity. In this scheme, quantum parameters become momentum-dependent, allowing the interior to be reworked and extended to the full spacetime. The resulting metric is found to be asymptotically flat, and its associated Kretschmann scalar remains regular throughout. Furthermore, it is shown that both the null expansion and Raychaudhuri equation are regular across the entire spacetime, indicating the resolution of the classical singularity.
The final chapters do not focus on a specific theory of quantum gravity but instead construct a class of time-dependent, asymptotically flat, spherically symmetric metrics to model gravitational collapse in quantum gravity. By imposing a quantum bounce, these metrics prevent singularity formation. They exhibit general properties expected from any quantum gravity theory, without committing to a particular approach. These metrics capture key insights into the dynamics of singularity resolution, as well as horizon formation and evaporation, following either a matter bounce or a black hole to white hole transition
It's Your Responsibility: The Socioeconomic Implications of Home-Based Renal Care
No full textMedical research has demonstrated that End-Stage Renal Disease (ESRD) is one of the most cost-consuming and labour-intensive chronic diseases. The frequent and perpetual need for dialysis treatments has created a crisis in renal care that has challenged health care systems to meet the current and future needs of the population. The crisis response in Ontario, Canada has relied upon the principles of neoliberalism to implement a home-based model of care that shifts dialysis from the hospital to the home. This transfer of treatment involves a significant downloading of work and costs to the patient, but offers them better health outcomes and autonomy over their care. To understand the socioeconomic outcomes of home dialysis programs, this dissertation draws upon interviews with patients, care partners, and frontline health care professionals to explore: 1) how home dialysis programs implement neoliberal processes of responsibilization which result in patients and their families performing their own care; and 2) how patients and their families respond to this responsibility within the context of their household. Major findings reveal that patients and their families experience significant hardships when transitioning to home-based care as they must negotiate divisions of labour within the family while managing the emotional and economic costs of treatment. In spite of these hardships, patients gain a significant amount of agency within the health care system upon their enrollment. Rather than being passive recipients of downloaded work and costs, they actively manage their care by directing the actions of frontline health care professionals, and influence wider care practices at the program level
The Changing Frameworks for Watershed Governance and Management in Ontario Considering Climate Change Effects
No full textThis paper analyses the evolution of provincial watershed governance in Ontario from the 1946 Conservation Authorities Act to the 2024 Provincial Policy Statement. Applying perspectives from historical institutionalism and Kingdon’s Multiple Streams Framework, this study examines the critical junctures, path dependence, and policy windows that defined the Ontario government’s historical approach to watershed governance and management. Twenty-six policies and laws were identified as the most significant developments in Ontario’s historical watershed governance framework, divided into five policy periods that reflect significant changes in government agendas and subsequent watershed policy direction. The analysis shows that beyond immediate environmental or health crises, the Ontario government’s approach to watershed governance is primarily dictated by shifting political agendas, rather than the constant presence of environmental pressures. The paper concludes with a call for provincial policymakers to consider the impacts of policies on Ontario’s essential watersheds as Doug Ford assumes his third term in office
Work Life Experiences of Black Women Leaders in Canadian Labour Union Organizations
No full textThe work life experiences of Black women leaders in Canadian labour union organizations (unions) are highly underresearched. Though, narratives promoted by unions indicating that unions are built on, and espouse values of, equity and fairness, fight for the rights of all their members across the table to employers and champion the rights of equity seeking groups suggest that these same principles are relevant to the experiences that Black women union leaders endure in their roles. Using the qualitative research method of phenomenology, and theoretical frameworks of critical race theory and intersectionality and everyday racism, the experiences of twenty-three Black women leaders who are currently active or were previously active (now retired) in paid or voluntary roles in Canadian unions were examined to produce knowledge about the distinctive work life experiences of Black women leaders in Canadian unions in order to make this information visible and widely available. Research findings reveal a myriad of ways that structures, policies, procedures, processes and practices disadvantage Black women union leaders primarily based on their race, and result in extremely poor, trauma-inducing and life-changing work life experiences with impacts that extend beyond the workplace
Twenty-Year Trends in the Prevalence and Predictors of Healthcare Provider Advice to Lose Weight: U.S. NHANES, 1999-2018
No full textHealthcare providers (HCP) play a critical role in screening for obesity and supports for obesity management. We examined 1999–2018 trends in the prevalence and predictors of U.S. adults being told to lose weight using ten nationally representative NHANES cycles (n=16,424). Survey-weighted descriptive statistics and multivariable logistic regression identified sociodemographic, anthropometric (BMI, WC), clinical (MetS, EOSS), and behavioral predictors stratified by age, sex, and ethnicity. Overall, 34.4% reported ever receiving weight-loss advice; the prevalence rose from 11.1% to 19.2% among overweight and from 20.9% to 33.4% among obese individuals. Adjusted models showed older age, female sex, higher BMI (OR overweight = 6.15; obese = 41.86), elevated WC (OR = 8.37), and EOSS stages 2 (OR = 3.90) and 3 (OR = 5.07) were strong independent predictors, while socioeconomic and behavioral factors were modest. HCP weight-loss advice increased over two decades but remains predominantly driven by simple anthropometric measures; integrating comprehensive clinical risk tools like EOSS may improve equitable obesity care
AdapTrain: Adaptive Model Partitioning for Efficient Independent Subnet Training on Heterogeneous and Dynamic Cloud Infrastructures
No full textModern distributed training systems face significant challenges in heterogeneous computing environments, where heterogeneity in computational resources among workers often leads to resource underutilization and extended training durations, particularly in resource-constrained environments. To address these challenges, we propose Adaptive Model Partitioning for Efficient Independent Subnet Training on Heterogeneous and Dynamic Cloud Infrastructures (AdapTrain), a novel framework that dynamically adjusts model partitioning to align with the computational capacities of heterogeneous workers. AdapTrain reduces the overhead of synchronization, thereby minimizing total end-to-end training time by ensuring synchronized completion of training rounds across all workers. Its adaptive design enables robust performance under workload variations, inherent resource heterogeneity, and multi-tenancy effects prevalent in cloud computing environments. An experimental evaluation of production workloads reveals that AdapTrain accelerates model convergence by more than 8x compared to the current training methods. Furthermore, AdapTrain integrates seamlessly into existing systems, introducing negligible system performance overhead while significantly enhancing training efficiency
Evaluating and Enhancing LLMs for Deep Learning Code Generation with DL-Bench
No full textLarge Language Models (LLMs) have recently demonstrated remarkable capabilities in automated code generation, yet their performance on domain-specific tasks such as deep learning (DL) pipelines remains underexplored. This thesis addresses this gap by introducing DL-Bench, the first comprehensive benchmark dedicated to evaluating LLMs on DL-specific code generation. DL-Bench comprises 520 carefully curated function-level tasks spanning all stages of the machine learning workflow, including data pre- and post-processing, model construction, training, inference, and evaluation, and is systematically categorized by pipeline stage, task type, and input modality. This fine-grained design enables detailed performance analysis and exposes unique challenges of DL code generation, such as tensor shape mismatches, framework-specific errors, and brittle reliance on phrasing.
Building on this benchmark, we further investigate robustness strategies for LLMs by proposing a prompt mutation pipeline combined with dual execution agreement. The pipeline systematically generates semantically equivalent prompt variations through lexical, grammatical, and naming transformations, which are then paired with model-generated test cases to diversify candidate solutions. Using a dual agreement framework, correct solutions are identified by their consistent success across test suites, mitigating common misinterpretations. To validate this approach, we evaluate three state-of-the-art LLMs, O4-Mini, DeepSeek R1 Basic, and Gemini 2.5 Pro, exclusively on DL-Bench. Results show that while baseline performance on DL-Bench is substantially lower than on general-purpose benchmarks, prompt mutations consistently yield measurable improvements (up to +2.9% pass@1), demonstrating their value in uncovering alternative correct solutions.
Overall, this thesis makes three key contributions: (i) the release of DL-Bench as a domain-specific, fine-grained benchmark for DL code generation, (ii) a systematic analysis of LLM weaknesses in DL contexts supported by a taxonomy of mutation effects, and (iii) the design and evaluation of a mutation-based dual agreement framework that enhances LLM reliability. These contributions provide both practical evaluation tools and methodological insights for advancing LLMs in specialized scientific programming domains. Future directions include scaling DL-Bench with multi-modal tasks, maintaining it as a live benchmark to track recency effects, and incorporating broader metrics such as code efficiency and maintainability
The Development of Parents of Children with Social, Emotional and Behavioural Difficulties Through Circle of Security-Parenting
No full textCircle of Security Parenting (COS-P) is an attachment-based psychoeducational intervention designed to enhance parents' ability to understand and respond to their children’s emotional needs and to foster secure parent-child relationships. This study explored the development of parents of children with social, emotional and behavioural difficulties through participation in COS-P. Using a mixed-methods design, the study focused on changes in parents’ internal representations of their children, emotion regulation, parental self-efficacy, and responses to their children’s negative emotions at pre-intervention, post-intervention and 3-months follow-up. The results indicated that following participation in COS-P, on average, parents developed more coherent internal representations of their children and the parent-child relationship. Additionally, meaningful increases in parental-self-efficacy, and reductions in self-reported unsupportive responses to children’s negative emotions were observed. Small improvements in emotion regulation were reported by some participants. However, there was also considerable variability in individual trajectories over time, highlighting the importance of employing both variable-centered and person-centered analyses. This dual approach provided a more nuanced understanding of the diverse ways in which parents responded to the intervention, highlighting how aggregated results may obscure important individual differences. The findings underscored the potential of COS-P as an effective attachment-based intervention capable of facilitating meaningful changes in parenting perceptions of themselves and their children even within a brief intervention period. Furthermore, the study demonstrated the feasibility and utility of the Five-Minute Speech Sample – Coherence (FMSS-C) as a low-burden, cost-effective tool for measuring changes in parents’ internal representations in clinical settings. These results contributed to the growing evidence supporting the use of attachment-focused interventions to enhance parent-child relationships
Tuning Big Data Systems Via Deep Learning
No full textModern database systems, including IBM Db2 have numerous parameters, “knobs,” that require precise configuration to achieve optimal workload performance. Even for experts, manually “tuning” these knobs is a challenging process. We present Db2une, an automatic query-aware tuning system that leverages deep learning to maximize performance while minimizing resource usage. Db2une uses a specialized transformer-based query-embedding pipeline and graph neural networks to feed as input to a stability-oriented deep reinforcement learning model. In Db2une, we introduce a multi-phased, database meta-data driven training approach—which incorporates cost estimates, interpolation of these costs, and database statistics—to efficiently discover optimal tuning configurations without the need to execute queries. Thus, our model scales to large workloads where executing queries repeatedly would be prohibitively expensive. Through experimental evaluation, we demonstrate Db2une’s efficiency and effectiveness over a variety of workloads. We show that Db2une provides recommendations surpassing those of other state-of-the-art systems and IBM experts