21793 research outputs found
Sort by
Prediction of Detonation Cell Size and Modulation of its Regularity in Gaseous Systems
No full textA detonation is a self-sustained, supersonic, combustion-driven compression wave which causes significant pressure and temperature changes. This phenomenon is relevant to the safety of engineering applications and industrial processes, as well as to the development of aerospace propulsion systems. A detonation wave typically displays a complex, nonlinear and unstable structure. This work aims to better quantify and predict characteristic length scales of the detonation structure, to clarify the influence of cellular regularity on detonation dynamics and to explore ways to modulate it. To that effect, the first half of this work is focused on developing a series of Artificial Neural Networks (ANN) using different chemical kinetic and thermodynamic input parameters to predict two main characteristic length scales, the detonation cell size and the critical tube diameter. The feedforward neural networks are trained and validated using available experimental data from the Caltech detonation database, covering a wide variety of gaseous combustible mixtures at different initial conditions. The second half of this work uses one- and two-dimensional numerical simulations to study the effect of O3 as an additive in the reactive mixture to study how it can change the instability and cellular structure of detonations. The use of microplates to modulate cellular regularity is also explored, with the modulated detonation used in the critical tube diameter problem in order to conclusively demonstrate the role of cellular instabilities on detonation dynamics and critical phenomena
Tramba: A Hybrid Architecture for Table Understanding
No full textThe increasing complexity and density of document images—particularly in scientific and industrial contexts—have posed significant challenges for traditional transformer-based models, due to their quadratic attention complexity and reliance on extensive computational resources. In response, this thesis proposes a novel hybrid vision architecture that integrates the Vision Mamba encoder with the Detection Transformer (DETR) framework to address the tasks of table detection and structure recognition. Leveraging Mamba’s state space modeling, which reduces computational complexity from O(N^2) to O(N), the proposed architecture retains competitive representational power while improving scalability and training efficiency. Vision Mamba is a state space sequence model designed for vision tasks, offering linear-time computation and efficient long-range dependency modeling through a bidirectional convolutional structure. DETR, in contrast, is an end-to-end object detection framework that formulates detection as a direct set prediction problem using a transformer-based encoder-decoder and learnable object queries. In our hybrid model, we replace DETR’s standard transformer encoder with a Mamba-based encoder stack, preserving the core object query mechanism while enabling lightweight and efficient sequential processing. Through extensive experiments on the PubTables-1M dataset, which is one of the largest datasets for table extraction tasks, we demonstrate that our model outperforms Faster R-CNN on both detection and structure recognition tasks, and approaches the performance of full DETR models—despite using only one-third of the encoder-decoder layers and fewer training epochs. These results highlight the architecture’s efficiency and adaptability, offering strong performance under constrained training budgets. Beyond empirical gains, the modular design of the model facilitates extensibility, including integration with large language models (LLMs) for advanced multimodal tasks such as document question answering, layout-based information retrieval, and regulatory content parsing. Finally, the lightweight nature of the Mamba encoder makes the model well-suited for deployment in enterprise-scale document processing systems, where throughput and latency are critical. This thesis thus introduces a promising direction for rethinking vision transformers through hardware-efficient sequence modeling, contributing meaningfully to the advancement of document AI and structured visual understanding. The source code is available at: github.com/SayeedAbid/Tramb
Exploring Therapeutic Factors: A Philosophical Inquiry into Art Therapy's Role in Healing Attachment Insecurity in an Adult Population
No full textAttachment insecurity in adulthood has been linked to emotion dysregulation, maladaptive relational patterns, and vulnerability to psychopathology. While several attachment-informed therapeutic models exist, few are tailored to address attachment insecurity through nonverbal, sensory-based modalities. This thesis examines the therapeutic potential of art therapy for addressing attachment insecurity in adulthood, employing philosophical inquiry to synthesize theoretical and empirical literature. Drawing on attachment theory, neurobiology, and art therapy frameworks, three core therapeutic factors are examined: sensory and tactile engagement, imagery and concretization, and therapist attunement within the triangular relationship. Findings suggest that the multisensory, embodied, and relational nature of art therapy aligns closely with the mechanisms of attachment formation and regulation, offering an alternative pathway for clients who struggle with traditional talk therapies. These therapeutic factors support emotional regulation, promote reflective functioning, and provide corrective emotional experiences, particularly for individuals who are high in attachment avoidance or have limited verbal access to emotional material. This study highlights the relevance of art therapy as an attachment-informed modality, offering a Tripartite Mechanism Model for Attachment Repair in Art Therapy, and advocating for its further integration and empirical investigation within adult mental health treatment
Tracing the Spectres, Evoking Absence-Presence: Memory-Making and Spatial Storytelling in Belfast City Centre and Lahore
No full textThis dissertation examines the processes of memory-making and spatial storytelling in Belfast City Centre and Lahore – two cities marked by sectarian violence, colonial legacies, and urban transformation. While memorials often provide fixed narratives of the past, this study explores more fluid and affective forms of memory that emerge through spatial storytelling, walking methods, and narrative interventions. Rather than adopting a conventional comparative framework, the research employs juxtaposition to highlight relational resonances between the two cities, tracing how memories of conflict surface in everyday urban life through absences, erasures, and spectral presences, or ‘melancholy survivals’. Using a mixed-methods approach that integrates walking go-along interviews with place-based professionals, analyses of planning texts, and literary narratives, the study interrogates how urban memory is shaped in the aftermaths of the Troubles in Belfast City Centre and Partition in Lahore. It considers how ‘post-conflict’ redevelopment and planning discourses often foreclose memory-making, while alternative modes – such as embodied movement, place-based storytelling, and literary texts – offer dynamic ways of engaging the past. Drawing on critical urban studies and memory studies, this research demonstrates that cities recovering from spatial trauma are shaped not only by material traces but also by narrative and affective interventions. Ultimately, the dissertation challenges fixed conceptions of urban heritage, emphasizing that memory-making in cities is an ongoing, contested, and deeply spatialized process
Exploring The Effects of Collectability and Scarcity Cues on Collectible Products Consumption
No full textScarcity tactics are widely used in marketing to enhance product desirability, yet their
effectiveness in the context of collectible products remains underexplored. While prior research
has established that scarcity can drive demand, few studies have examined how these tactics
interact with perceptions of collectability, particularly within specific markets such as vinyl
record collecting. My research addresses this gap by investigating how scarcity and collectability
cues influence consumer responses to collectible products. In my research, I focus on vinyl
records due to their recent resurgence and cultural significance among music collectors. I
explored the interplay between collectability and scarcity cues across four studies. A pilot study
and a pre-test first helped design relevant manipulations. Two experimental studies then
examined the effects of collectability (e.g., special features) and scarcity (e.g., limited quantity)
cues on consumers’ responses (i.e., ownership desire, purchase intentions, and anticipated
regret). These studies also explored the roles of the perceived economic and emotional values of
the collectible products as potential mediators. My preliminary findings suggest that scarcity and
collectability cues function as distinct drivers of consumer responses, as only main effects were
found. Additionally, the perceived economic and, to a lesser extent, emotional value of the
collectible product mediated these effects. My research contributes to the literature on scarcity
marketing tactics by exploring how they may impact collectible product consumption, and to the
literature on collectible products, by experimentally testing the effects of collectability cues. My
findings also offer insights for marketers seeking to leverage collectability and scarcity cues
within collectibles markets
Design of Novel Rare-Earth Cluster-Based Metal–Organic Frameworks
No full textNearly thirty years after the first use of the name metal–organic framework (MOF), more than 50,000 different non-disordered MOF structures have so far been reported, according to the Cambridge Structural Database (CSD). Yet, the field still holds immense opportunities for investigation into the nearly infinite variety of linker and metal combinations. These organic linkers and metal nodes combine to form framework structures, which are often three-dimensional, crystalline, and display extensive porosity. Moreover, by carefully designing MOFs by selecting linkers, metals, and targeting specific frameworks, different applications can be envisioned, including gas storage, drug delivery, chemical separations, light-harvesting and energy conversion, catalysis, sensing, and adsorption. Many of these applications take advantage of open metal sites, where the Lewis acidic metals are involved in host-guest interactions and processes that benefit from an ordered site that can accept an electron pair.
Due to the special characteristics of rare-earth (RE) elements, which include scandium, yttrium, and the whole series of lanthanoids, promising new RE-based MOFs have been obtained in the past years. Among these characteristics, it is worth mentioning the high coordination numbers and distinct optical properties of RE(III) ions, which can lead to the generation of materials with interesting photophysical and photochemical properties and unique crystalline structures, potentially featuring stable and accessible open metal sites.
In this work, a series of RE-MOFs based on the archetypical zirconium MOF-808 was obtained through a de novo synthetic approach. These MOFs were named as RE-CU-45 (CU = Concordia University) and feature six-connected hexanuclear RE(III)-clusters bridged by 1,3,5-benzenetricarboxylic acid linkers. Studies were carried out to optimize the synthetic conditions used to obtain the new RE-MOFs, mainly exploring the metal precursors that are used, focusing on a balance between high purity, yield, suitable crystallite sizes, and reproducibility of the synthesis. Additionally, the RE-MOF has been tested in regard to the possibility of accessing its pores and open metal sites through a series of post-synthetic modification by solvent assisted ligand incorporation. The final materials are fully characterized by powder X-ray diffraction (PXRD), N2 sorption, thermogravimetric analysis (TGA), diffuse reflectance infrared Fourier Transform spectroscopy (DRIFTS), scanning electron microscopy (SEM), nuclear magnetic resonance (NMR) spectroscopy, and studied regarding their photophysical properties through diffuse reflectance UV-vis spectroscopy (DR-UV-vis) and photoluminescence spectroscopy
The Art of Teaching Play: A Phenomenology of Froebel’s Gifts in Practice
No full textThis research explores the art and education history of Kindergarten gift play, a guided activity with abstract blocks and learning objects. Through a phenomenological inquiry, it asks: what is the essence of teaching through play with the Froebel Gifts? Its presentation of Gift Play object-interviews with former teachers at the Froebel Education Centre in Mississauga, Ontario, encounter both abstract and concrete constructions: first, Kindergarten’s 19th-century educational beliefs and practices as first introduced by its German founder, educational reformer Friedrich Froebel (1782–1852); and second, the concrete and symbolic Kindergarten system of Gifts and Occupations, objects and guided activities for learning through play. Hermeneutic phenomenology and posthumanist object-interview heuristics structure a holistic, multi-dimensional, and material-sensitive description of the ‘essence’ of ‘teaching play’ and ‘gifts’ as lived and as living, reanimated in renewed understandings of Froebel’s gifts, to pedagogy, and to the art of play
Using Latent Profile Analyses to Differentiate Callous-Unemotional Trait Profiles In a Community Sample of Youths
No full textResearchers have begun to distinguish between primary and secondary profiles of callous-unemotional (CU) traits among youths. The proposed etiological pathway of primary CU traits is often theorized to stem from temperamental or biological deficits in processing emotional cues, while secondary CU traits is often theorized to stem from response to trauma or maltreatment. However, there is no consensus on the best measurement of CU profiles. This study compared different measurements of CU profiles using latent profile analyses. Participants were 243 youths (Mage = 12.07, SD = .91; 44.86% female), drawn from a larger study examining the development of risky behaviours (Collado et al., 2014; MacPherson et al., 2010). Latent profile analyses were conducted to identify subgroups of youths using different combinations of CU traits, anxiety, exposure to maltreatment/trauma, and emotion dysregulation as indicators. I hypothesized three groups: high primary CU, high secondary CU, and low symptoms. I also hypothesized that CU and anxiety indicators would be best at differentiating profiles. Results indicated two models with good fit: a model using CU traits, anxiety, exposure to trauma, and emotion dysregulation as indicators (AIC = 5536.940, BIC = 5617.281, SABIC = 5544.374, Entropy = .869) and a model with CU traits, anxiety, and exposure to trauma as indicators (AIC = 4017.569, BIC = 408.444, SABIC = 4023.387, Entropy = .870). Both models yielded four groups: low symptoms (68.31% and 69.96% respectively), high trauma (9.47% and 9.05%), high anxiety (17.70% and 16.46%), and secondary CU (4.53% and 4.53%). No model identified a primary CU profile. Results inform researchers and clinicians of important features (i.e., CU traits, anxiety, exposure to trauma, and emotion dysregulation) to identify youths with secondary CU traits and, consequently, those at higher risk for persistent disruptive behaviours in a community sample
Feature-Centric Approaches to Non-Intrusive Load Monitoring and Appliance Identification
No full textLoad disaggregation refers to estimating appliance-level consumption from overall household energy data. It includes tasks like load identification and energy disaggregation. Researchers are actively developing various machine learning and deep learning techniques to disaggregate total household energy consumption into appliance-level usage. At the same time, many are focusing on identifying individual appliance loads to detect faulty devices or to improve the overall disaggregation process. This thesis makes two significant contributions to the field, addressing the challenges of total load separation and appliance identification. The first contribution focuses on energy disaggregation using a simplified Feed-Forward Neural Network architecture optimized for performance and efficiency. Oversampling techniques are developed for training data to improve the detection of
appliance activation cycles. Furthermore, the model incorporates additional features derived from aggregate consumption profiles, enhancing input diversity and robustness. This approach is tested on the RAE, REFIT, and REDD datasets under both clean and noisy conditions. The second contribution addresses appliance-level load identification using a Kolmogorov–Arnold Network, offering a lightweight and efficient alternative to deep models. Around 75 features are extracted from voltage and current signals, grouped into statistical, power-related, and frequency-domain categories. An effective feature selection process is conducted using multiple tests and correlation matrices to retain only the most informative inputs, thereby reducing model complexity and enhancing generalization. Additionally, we tune the hyperparameters of the KAN to control the degree of oversampling, allowing it to better handle imbalanced data. The model is evaluated using three public datasets: COOLL, PLAID, and WHITED
Structural Inequality in Workload Allocation among Gig Drivers: Residency Status, Language Proficiency, and Human Capital in Quebec’s Last-Mile E-commerce Delivery Sector
No full textE-commerce’s rapid growth has intensified demand for last-mile delivery. In response, many firms have shifted from direct employment to multi-tiered outsourcing models that rely heavily on gig workers. In Quebec’s bilingual labour market, this fissured workforce can amplify inequities in work allocation. We examine how residency status (permanent resident, study visa, work visa) and proficiency in French and English shape weekly and daily workload allocation and delivery performance (delivery success and customer complaint rates). We also test whether returns to human capital—job-specific skill capital (prior parcel-delivery experience), job-specific physical capital (vehicle-ownership tenure), and general human capital (driver age)—vary by residency status and language proficiency.
Drawing on operational records from a last-mile delivery company in Montreal, we analyse 129 drivers: 56.2% hold work visas, 22.6% are permanent residents, and 21.3% hold study visas; none report French or English as a first language. Most report intermediate French proficiency (64%), followed by advanced French (16%), beginner French (10%), and intermediate English (10%).
Findings show persistent structural inequalities in delivery workload allocation: drivers on work visas received significantly fewer weekly and daily allocations than permanent residents, while study-visa drivers do not differ on average. Lower language proficiency consistently reduces assigned workloads. Prior delivery experience and longer vehicle ownership increase allocations, whereas age has no main effect. Returns to human capital are heterogeneous—larger for study-visa drivers but attenuated or even negative at low language proficiency. Delivery performance mirrors these patterns: study-visa drivers exhibit slightly lower success rates, and limited-proficiency drivers face higher complaint rates, particularly as experience rises.
Overall, residency status and language proficiency generate structural inequalities in delivery work allocation and human capital only partially offsets. Managerially, language training, same-language mentoring, fairness-aware dispatch algorithms, and skill-matched route allocations can mitigate disparities. Policy recommendations include incorporating fairness principles in contracts, facilitating preferred-language use at work, and expanding support for skill development