Recherche uO, le dépôt numérique de l'Université d'Ottawa, réunit le matériel de recherche et d'enseignement créé par notre communauté universitaire et nos partenaires. Le savoir de l'Université est ainsi disponible à long terme et en accès libre, ce qui lui procure de la visibilité et facilite sa diffusion.

Nouveaux dépôts

  • Type d'Item : Item ,
    DQN-Guided Rule-Class Selection for Search-Based Query Rewriting
    (Université d'Ottawa / University of Ottawa, 2026-09-16) Liu, Shengchen; Kantere, Verena
    Query rewrite transforms a SQL query into a semantically equivalent form expected to execute more efficiently. Search-based rewriters such as LearnedRewrite organize alternative rewrite orders as a policy tree and explore it with Monte Carlo Tree Search. In the publicly released LearnedRewrite implementation, expansion enumerates every applicable rule class at each search node, and each candidate triggers a rewrite execution and a PostgreSQL EXPLAIN call, making expansion a measurable contributor to rewrite latency. This thesis formulates expansion-time rule-class selection as a state-dependent decision problem and proposes DQRS (DQN-Guided Rule-Class Selection), which ranks the legal rule classes at each state with a Deep Q-Network and forwards only the top k to the unchanged rewriter routine. On an in-domain workload derived from the TPC-H benchmark at k = 1, average rewrite latency falls from 15.3 ms to 4.0 ms and average policy-tree size from 9.9 to 2.4 nodes, reductions of about 74% and 76%. Average cost reduction is preserved while the median falls from 156 to 103, so the method exposes an explicit operating point on the efficiency–quality trade-off rather than dominating the baseline. Under an identical legal-action mask and at the same node count, learned ranking attains a median cost reduction of 103 against 13 for random selection. Transferred without retraining to a workload derived from the TPC-DS benchmark at k = 4, DQRS reduces latency by about 24% and node count by about 30% at unchanged cost reduction, but in this saturated regime it is not distinguishable from random selection. The baseline throughout is the released implementation, which omits the learned cost estimator of the original paper, and all quality figures are optimizer-side estimates rather than measured execution time. Extending the evaluation from optimizer estimates to end-to-end execution time is the primary direction for future work.
  • Type d'Item : Item ,
    Comparative Analysis of Computational Methods for Optimal Transport
    (Université d'Ottawa / University of Ottawa, 2026-09-16) Zhytkevych, Ivan; Zarepour, Mahmoud
    This thesis presents a comparative and experimental study of computational methods for discrete optimal transport under the squared-Euclidean cost. We benchmark the exact simplex method, the Sinkhorn algorithm, first-order and quasi-Newton dual ascent, a chi-square-regularized primal-dual hybrid gradient (PDHG) method, and the Back-and-Forth method, across dimensions d in {1, 2, 3}, a range of grid resolutions and regularization levels, and a family of distributions varying in modality, tail thickness, skewness, and anisotropy. Solvers are assessed against three criteria - stability, runtime, and accuracy - and the conclusions are validated on a colour-transfer application. There we find that in RGB colour space the exact solver is not always preferable to a smoother, slightly biased map, which scores better on the structure-preservation metrics of SSIM, gradient correlation, and Laplacian sharpness. Beyond the comparison, we introduce a new method for the outer-regularized 2-Wasserstein barycenter based on Sobolev gradient ascent on the dual functionals combined with the Back-and-Forth c-transform, the simplest case of a broader Back-and-Forth Flow framework.
  • Type d'Item : Item ,
    Memorias del Enemigo: Relatos Autobiográficos de Heberto Padilla, Reinaldo Arenas y Eliseo Alberto sobre el Poder en la Cuba Revolucionaria
    (Université d'Ottawa / University of Ottawa, 2026-09-16) Garcia Rodriguez, Carlos; Guerrero, Jorge Carlos
    This dissertation examines the totalitarian drift of the Cuban Revolution through a critical reading of La mala memoria (1989) by Heberto Padilla, Antes que anochezca (1992) by Reinaldo Arenas, and Informe contra mí mismo (1997) by Eliseo Alberto. From an interdisciplinary perspective that integrates literary study with political analysis, I demonstrate how these autobiographical works reveal the mechanisms of censorship, surveillance, repression, exclusion, and control deployed on the island after 1959. This theoretical approach is fundamentally inspired by a contextualization of Carl Schmitt's political thought, particularly regarding the distinction between friends and enemies that radically shaped Cuban society. In the chapter dedicated to La mala memoria, I explain how the confessional nature of these memoirs is permeated by the narrator's guilt complex in relation to the "Padilla affair," an episode where simulation is revealed as constitutive of totalitarianism. In addition to illustrating how Padilla outlines the sacrificial orientation of revolutionary History, I complete the section with his literary portrait of Fidel Castro, whose tyrannical representation appears in tension with the author's self-portrait. In my initial approach to the study of Antes que anochezca, I argue that Arenas's writing style can be conceived as a principle of individuation that challenges certain dismissals of his autobiography. After unveiling the castrating logic of Castroism described by Arenas, I analyze the manifestations of collective violence that gave rise to what I define as a "dejective biopolitics". In the chapter on Informe contra mí mismo, drawing on Eliseo Alberto's experience, I problematize the confession and guilt stemming from his collaboration with state surveillance apparatuses entrenched in a permanent state of exception. After integrating various critical recollections of the Revolution - presented as letters, slogans, narratives, and even the portrait depiction of Castro created by Eliseo Alberto, I elucidate the progressive erosion of the revolutionary myth following the collapse of the socialist bloc. In sum, through the autobiographical texts of these exiled writers, I demonstrate the presence of a literary corpus that challenges the hegemonic, apologetic, and totalizing narratives of the revolutionary process.
  • Type d'Item : Item ,
    Chip-Integrated Nanowire Quantum Dot Emitters for Scalable Quantum Photonics
    (Université d'Ottawa / University of Ottawa, 2026-09-16) Yeung, Edith Pui Cheung; Dalacu, Dan
    Single-photons are a fundamental component in many quantum optical technologies. An example of where single-photons are needed is in quantum cryptographic technologies such as the BB84 quantum key distribution protocol. This involves encoding data in the polarization of single-photons to share a key between the two parties. Such applications require a single-photon source that ideally can emit exactly one photon on-demand with a well-defined polarization and spatiotemporal mode. Among the various candidates, semiconductor quantum dots are two-level systems that can be tailored to emit single-photons at desired wavelengths. However, since the quantum dots emit randomly in all directions, they are typically embedded into different photonic structures for more efficient coupling to optical fibre systems. In particular, bottom-up III-V nanowires embedded with quantum dots are promising single-photon sources for integration with large-scale quantum photonic circuits. To test the feasibility of a nanowire-Si₃N₄ waveguide integration, simulations of the power transfer from the fundamental HE₁₁ mode of a tapered nanowire to an underlying ridge waveguide predicted >95% coupling efficiency. The mode within the nanowire is expanded as the light propagates up the nanowire taper, making it compatible for both collection with a free-space objective or to evanescently couple to an in-plane straight waveguide. In this thesis, different hybrid quantum photonic integrated circuit devices are demonstrated to efficiently harvest single-photons generated on-chip. Through a pick-and-place method, individual nanowires can be transferred from the growth substrate to a chip containing Si₃N₄ photonic waveguides. This hybrid integrated device combines the efficient emission of quantum dots with the low-loss characteristics of waveguides. Coupling light on and off the chip is performed using lensed fibres. An evanescent coupling efficiency of η_c ~ 84.8% was measured with one device. Indistinguishable photons were successfully generated on-chip using above-band excitation and quasi-resonant (p-shell) excitation to obtain two-photon visibilities of 9.2% and a much improved 19.2%, respectively, over the temporal extent of the emitted photons. A major limitation of these devices is that single-photons are more efficiently collected from the tapered end of the nanowire as opposed to the base. To collect the photons emitted in both directions, a curved waveguide with a long untapered nanowire can be used. This requires the precise placement of the quantum dot adjacent to the curved waveguide. To increase the tolerance for the quantum dot position along the curve, a curved segmented waveguide was investigated. A coupling efficiency of 63% was obtained from the combined counts of both the X and X⁻ transitions in a single dot. For large-scale quantum networks, stable single-photon sources with a small footprint are desirable. The need for active alignment is eliminated with a plug-and-play source which allows for easy access to multiple fibre-coupled sources simultaneously. A fibre-packaged chip consisting of straight waveguides and tapered nanowire-quantum dots is demonstrated to have a total plug-and-play efficiency of 3.6 ± 0.7% simultaneously collecting from both ends of one device. The multiphoton emission probability of the device is g⁽²⁾(0) = 7.2%. The fibre-packaged chip was robust against multiple cooldown cycles and showed stable alignment over 27 hours of continuous operation. The devices demonstrated in this thesis serve as proof-of-concept platforms for in-plane single-photon generation. Based on this work, single photons can then be manipulated with standard photonic components including ring-resonators, Mach-Zehnder interferometers, phase shifters for more complex applications on-chip.
  • Type d'Item : Item ,
    Exercise Training for Wheelchair Users Living with Chronic Neurological Conditions: A Summary of Current Evidence and Future Research Directions
    (Université d'Ottawa / University of Ottawa, 2026-09-16) Liu, Sharon; Pilutti, Lara A.
    Exercise training has emerged as a beneficial approach to chronic disease management for individuals living with chronic neurological conditions (CNCs). A variety of challenges (e.g., complex impairments, physiological deconditioning, environmental accessibility) may impact exercise engagement for individuals with CNCs at higher disability levels. Understanding the current landscape of the literature may help advance exercise research and implementation for this group. This thesis aimed to identify and summarize outcomes and exercise prescriptions in exercise training interventions for people living with CNCs who use wheelchairs for mobility. Five electronic databases were searched: Medline, Embase, CINAHL, SPORTDiscus, and Scopus for articles published up to May 2025. The search strategy incorporated terms related to 'chronic neurological conditions', 'wheelchair use', and 'exercise'. Outcomes were mapped to the International Classification of Functioning, Disability, and Health (ICF) Framework and compared to CNC-specific Brief ICF Core Sets. Exercise prescriptions were organized according to Frequency, Intensity, Time, and Type (FITT) components and compared to CNC-specific exercise recommendations. 176 articles were identified for inclusion. The most frequently identified ICF category was body functions (n=1613, 72%), followed by body structures (n=393, 18%). MS studies had the most outcome overlap (47%) between extracted outcomes and the ICF Core Set for MS. The most frequently reported FITT component was exercise type (100% exercise prescriptions), whereas exercise intensity was poorly reported (40% of exercise prescriptions). Only two (8%) exercise prescriptions met recommended targets overall. Future studies of exercise training should prioritize incorporating a range of outcomes with relevance to individuals living with CNCs, while also considering study design and sample characteristics. Improved reporting of FITT elements, particularly exercise intensity is needed. Addressing the disconnect between what is tested in the research setting and core outcome sets and exercise recommendations for use by public and healthcare providers should be considered in future work.