39456 research outputs found
Sort by
Deep Generative Models for Trajectory Prediction and Mobility Network Forecasting
No full textPredicting human mobility is essential for urban planning, traffic management, and epidemiology. This thesis tackles two intertwined challenges: accurately forecasting individual trajectories and inferring the resulting mobility network. First, we introduce TrajLearn, a Transformer‑based deep generative model that treats trajectories as token sequences and employs spatially constrained beam search to predict each individuals’s next k locations with high precision. Building on these forecasts, we present MobiNetForecast, which constructs and predicts the future topology of the mobility network by detecting when independently predicted trajectories intersect in space and time. Across large, real‑world datasets, our unified framework achieves up to 40% relative gains in trajectory accuracy and up to 100x improvement in contact prediction over state-of-the-art baselines. These results demonstrate that combining advanced sequence modeling with explicit contact inference offers a powerful, scalable solution for dynamic mobility network forecasting
The ‘Visa Student Dream’: An Examination of Shifting Trends and Vulnerabilities in Chinese International Student Populations Within Toronto’s Secondary Schools
No full textThis study aims to explore the experiences of Chinese international students in secondary schools in Canada, paying particular attention to the vulnerabilities and risks among this population. Along with the pre-existing systemic issues international students face, the COVID-19 pandemic placed these students in an even more precarious position due to the anti-Asian racism and discrimination that manifested itself during this time period. The study employed Critical Race Theory (CRT), specifically Asian Critical Theory (AsianCrit) which prioritizes Asian identity and their experiences with racism to understand and contextualize how prevailing systems of oppression impacted the lives of Chinese international students. Utilized alongside these theories are complementary frameworks like International Student Security (ISS) and neoliberalism to further explore the experiences, vulnerabilities and risks among this population. Multiple constructs were also used like model minority, yellow peril, neo-racism, and racial capitalism to expand understanding and application of theories such as CRT and AsianCrit to this international student population. The principles highlighted within CRT and AsianCrit theories, utilized alongside the frameworks and constructs all built upon each other to provide further insight into how educational institutions continue to operate within a dominant culture paradigm and with whiteness as a norm and how programs are maintained and/or implemented based on assumed notions and ideologies of Asian international students. Interview data was collected from six teachers and six Chinese international students from public secondary schools within the Greater Toronto Area (GTA). Findings revealed that neoliberalism has ultimately created unethical and unsafe policies and practices that negate EDI initiatives, and the push towards standardization and individualism perpetuate racial inequities. Standards of success are rooted in Whiteness that validates Western knowledge, with concepts like racial capitalism playing a role in exploiting Chinese international students. Issues and concerns regarding the academic, social, and housing/guardianship experiences of Chinese international youth were also revealed, as students faced discrepancies between the services and support that was advertised in comparison to what they actually experienced in Canada. Key concerns and gaps in policy and programmatic supports were also identified for international students with the study outlining recommendations and interventions to better support international secondary students during their studies in Canada
Graph Learning and Optimization for Irregular-Structured Signal Processing
No full textGraph Signal Processing (GSP) extends harmonic analysis tools, such as Fourier transforms and wavelets, to discrete signals defined on finite graphs, enabling tasks like signal denoising, prediction, and interpolation on irregular domains. A critical first step in GSP is to learn an appropriate graph that captures pairwise similarities or correlations inherent in the data, ensuring that subsequent graph-based filtering effectively leverages local structure for improved performance.
However, most existing graph learning methods assume static relationships, while real-world interactions often evolve over time. To address this problem, this thesis proposes a slowly time-varying graph learning framework that models the difference between consecutive adjacency matrices as a low-rank matrix. This approach accommodates gradual shifts in node-to-node similarities over time, enabling efficient graph updates with low computational overhead while maintaining alignment with the underlying data.
Beyond graph construction, the challenge of dense or complete graphs often arises, particularly in large-scale applications where representing all possible edges is computationally prohibitive. To address this issue, this thesis introduces a sparsification method guided by the Fiedler number, the second smallest eigenvalue of the Laplacian, which quantifies graph connectivity. By removing edges that minimally affect the Fiedler number, the resulting sparser graph preserves essential connectivity while significantly reducing training and inference costs for deep learning models (e.g, graph convolutional networks (GCNs)). Together, these contributions provide a flexible and computationally efficient approach to GSP in dynamic and large-scale graph settings
Photography and the Extractive Gaze: Visual culture and natural resource extraction on the Canadian Shield, 1900-1930
No full textWhile the Anthropocene is a relatively new concept, communities in extractive zones have already experienced the transformations, changes, and destruction brought by human activity and industrial development. Through archival research, my dissertation investigates how vernacular and survey photographs of mining from the early 20th century circulated to promote and occasionally challenge resource extraction. Taking as its case study the Timiskaming region on the Precambrian shield, I explore how photography chronicled the transformation of territory under extractive capitalism as photographs invited viewers to envision new futures, where commerce and art would come together to drive economic and industrial development. These archival photographs formed a site of knowledge production that shaped how people in the early 20th century understood industrial development. I situate contemporary experiences of climate change within these longer historical trajectories to trace the social and environmental legacies of resource extraction.
Once extracted from the earth, raw natural resources are transformed into consumer goods, which bear little evidence of the complex networks of human and non-human labour that brought them into being. Environmental art historians have addressed this disconnect by examining how human cultural production has unfolded within a broader ecological context, challenging the separation of art from the natural world. This dissertation makes a material link between silver and photography during a period where the demand for silver bullion exploded due to the rise of amateur photography. My research identifies connections between photographic technologies, visual form, and political activism. I conclude that photography can bring the often-invisible processes of extraction into view and document the historical production of environmental trauma. I read the histories that the archival photographs contain, and the histories that were foreclosed, as documents that can teach us something about hope, healing, and living in the Anthropocene
Assessing and Enhancing the Quality of News Headlines Using Machine Learning
No full textHeadlines play a pivotal role in capturing readers' attention, and their quality is critical for engaging audiences. In this thesis, we propose various solutions to assist news media in crafting high-quality headlines. First, we delve into headline quality assessment, devising four innovative indicators that automatically evaluate headlines' quality. Our proposed model empowers news outlets to automatically determine the quality of published headlines. We evaluate the quality of headlines from The Globe and Mail using these four indicators and provide insightful results. We then use this labeled data to train our novel headline quality prediction model to predict the quality of unpublished headlines, assisting journalists in selecting high-quality headlines for their articles. Furthermore, we facilitate journalists' work by recommending high-quality headlines for their articles. To accomplish this, we propose a headline generative model that learns to generate headlines using Reinforcement Learning (RL). Our model can be optimized not only with respect to a non-differentiable metric but also based on a combination of two different metrics simultaneously. Additionally, we enhance headline generation in terms of both training speed and the quality of the generated headlines by proposing a novel architecture utilizing state-of-the-art transformer models. In our architecture, after generating candidate headlines using state-of-the-art models, we select the most popular headline using our headline popularity prediction model. Moreover, we establish a popularity benchmark for evaluating headline generation models based on their ability to generate popular headlines. Lastly, we forecast changes in how people consume news articles, envisioning a shift towards interacting with agents instead of navigating news portals. To address existing challenges and enable this transition, we introduce Semantic In-Context Learning (S-ICL), an innovative approach enabling Large Language Models (LLMs) to deliver updated news in a conversational format, enhancing user engagement and comprehension for news media
The Impact of Entrepreneurship Education on the Development of Entrepreneurial Intention Among Engineering Students: The Mediating Role of Entrepreneurial Mindset and Self-Efficacy.
No full textThis thesis explores how entrepreneurship education (EE) influences the development of entrepreneurial intention (EI) among engineering students, specifically examining the mediating roles of entrepreneurial mindset (EM) and entrepreneurial self-efficacy (ESE). Considering the growing global importance of innovation and entrepreneurial skills, the study investigates the degree to which EE enhances EI by fostering EM and strengthening ESE. Using a quantitative survey of 431 engineering students at York University’s Lassonde School of Engineering, the study applied structural equation modeling to examine the direct and indirect effects of EE on EI through EM and ESE. Our findings show that EE significantly enhances EI, with both EM and ESE positively mediating this relationship. In this research, we investigated the impact of entrepreneurship education on entrepreneurial mindset and on entrepreneurial intention. In reality this relationship is more complex, and causality might be in the opposite direction. Future research should investigate the interplay between these entrepreneurial components and the iterative nature of their evolving relationships. This highlights the value of integrating EE into engineering curricula to develop the EM needed in today’s technology-driven world. The research contributes to existing literature by quantifying EE's impact on EI and offers practical implications for educational policy and curriculum development, advocating for the continued inclusion of EE to effectively prepare engineering students for entrepreneurial careers and foster economic innovation and growth
Mapping Child Marriage Within the History of International Law: A Turn to the Archives of the 1926 Slavery Convention
No full textAt its heart, this thesis is an investigation into how the female child, compelled into conditions akin to slavery through marriage, has been historically marginalized from the scope of slavery as it is defined in the 1926 Slavery Convention. Conversely, the Human Rights Council of the United Nations General Assembly recently recognized that the experiences and exploitation of women in forced marriages can meet the international legal definition of slavery. This framing evolution indicates the reorientation of institutional efforts in recognizing forced marriages as a form of slavery. Through a historical study of the international legal origin of slavery, this thesis probes into the conceptual and linguistic shift in the framing of forced marriages. In doing so, it identifies, within the archives of the League of Nations, the forces which shaped a narrow conception of slavery in the law.
A critical analysis of the intersections of law and gender during the colonial era of the League of Nations concludes with the identification of a systematic exclusion of child marriage from the legal construction of slavery, driven by hegemonic forces. This work finds that the politics and ideologies of coloniality shaped a narrow conception of slavery, enabling the continued economic, labor, and other forms of exploitation of the colonized Global South. It emphasizes the limitations of the prevailing anti-slavery framework, rooted in this history, which continues to relegate the enslaved child to the margins
Chronically Excluded? Public Toilet Access for Youth with Gastrointestinal Illnesses
No full textThis thesis explores the intersection between public infrastructure, health and youth geographies, time geography, and sensuous and emotional embodiments to highlight public toilets as a critical yet often overlooked urban space. Through these intersections, this thesis not only spotlights public toilets as central nodes in everyday life, but also the differential ways these spaces impact populations who rely on them most for medical needs. Through a feminist methodological approach that employs semi-structured interviews and space-time diaries, this thesis asks: How do the daily mobility patterns of youth with chronic gastrointestinal illnesses depend on the spatial and temporal availability and accessibility of public and private toilet facilities? This thesis investigates the constraints to mobility and wellness that these individuals face when met with inadequate and inaccessible toilet infrastructure, with a case study in the Greater Toronto Area. Encompassing both suburbs and city centre, the research sample illustrates the infrastructural disparities between dense and sparse landscapes. From the ‘in-betweens’ from one toilet to the next, to the sensuous and emotional experiences felt within these spaces themselves, this research investigates how the everyday lifeworlds of chronically ill youth – through work, school, and play – can be enabled and disabled by the quality of infrastructure they are met with, and the coping mechanisms they employ to aid their journeys and experiences, attributing to overall wellness
Application and Optimization of Prompt Engineering Techniques for Code Generation in Large Language Models
No full textLarge Language Models have demonstrated remarkable capabilities across various domains, particularly in code generation and task-oriented reasoning. However, their accuracy and reliability in generating correct solutions remain a challenge due to the lack of task-specific prior knowledge and the limitations of existing prompt engineering techniques. Current state-of-the-art approaches, such as PAL, rely on manually crafted prompts and examples but often produce suboptimal results. Additionally, while numerous prompt engineering techniques have been developed to improve performance, selecting the most effective technique for a given task remains difficult since different queries exhibit varying levels of complexity.
This work presents an integrated approach to enhance the application and optimization of prompt engineering for code generation. First, it introduces TITAN, a novel framework that refines language model reasoning and task execution through step-back and chain of thought prompting. TITAN eliminates the need for extensive manual task-specific instructions by leveraging analytical and code-generation capabilities, achieving state-of-the-art zero-shot performance in multiple tasks. Second, it proposes PET-Select, a prompt engineering agnostic model that classifies queries based on code complexity and dynamically selects the most suitable prompt engineering technique using contrastive learning. This approach enables Pet-Select to optimize prompt selection, leading to improved accuracy and significant reductions
in token usage.
Comprehensive evaluations across diverse benchmarks, including HumanEval, MBPP, and APPS, demonstrate the effectiveness of TITAN and Pet-Select. TITAN achieves up to 7.6 percent improvement over existing zero-shot methods, while Pet-Select enhances pass@1 accuracy by up to 1.9 percent and reduces token consumption by 49.9 percent. This work represents a significant advancement in optimizing prompt engineering for code generation in large language models, offering a robust and automated solution for improving performance in complex and diverse programming tasks
The Effect of Optic Flow During Attention Related Tasks and Quiet Stance
No full textVision plays a significant role in balance as it provides visual cues to help maintain balance. Increasing visual information during upright stance leads to tighter regulation of upright stance. Cognitive related tasks and postural control share capacities and compete for cognitive resources which may cause interference on one or both tasks. There is limited work on the impact of visual gain manipulation on dual tasks that include attention tasks that require visual information while maintaining an upright stance. While previous studies have explored how dual tasks and modified visual feedback influence balance separately, there is limited work on the impact of visual feedback during dual tasks that require visual information.
This study explored how visual cues affect balance when combining effects of visual conditions and different cognitive attention tasks. Optic flow was amplified or reduced to 4 gain conditions (0.25x,1x,4x,16x) within virtual reality (VR) relative to the head position while participants stood quietly on a force plate that measure ground reaction forces and completed 12 randomized trials across 3 conditions. Kinematics were collected through 8 markers placed on different parts using motion capture. Muscle activity was also collected by placing EMG on 3 lower leg muscles during the trials. MSRS questionnaire was completed after each trial to assess movement consciousness. Root mean square (RMS), and the mean power frequency (MPF) of Centre of Pressure (COP) and head position (HeadPos) were used to quantify balance. The mean angle RMS of relative angular displacement of the hip, knee and ankle was calculated, and the mean angle RMS of absolute angular displacement of the trunk, thigh, shank, foot was calculated to quantify balance. Developing a greater understanding of complex dynamics of visual feedback on cognitive task that requires vision while quiet standing, may enhance our understanding of how visual information aids postural control during dual tasks