21090 research outputs found
Sort by
Impact of short-acting vs. long-acting antipsychotic use on time in seclusion on a forensic assessment unit: a retrospective chart review
No full textBackground: Seclusion, a strategy used to manage aggressive behavior in patients posing safety risks, can result in prolonged hospitalization and trauma. Antipsychotics, used to treat major mental illnesses including schizophrenia and bipolar disorder, are available as long-acting (LA) and short-acting (SA) formulations, with evidence suggesting LA antipsychotics may improve patient outcomes such as reducing hospitalizations and decreasing aggression, which is primarily driven by an increase in medication adherence.
Objective & Methods: This retrospective chart review evaluated the impact of SA versus LA antipsychotics on seclusion duration in adult patients at the Southwest Centre for Forensic Mental Health Care (Ontario, Canada) between April 1, 2017, and December 31, 2023.
Results: Of 83 patients (60 in the SA cohort, 23 in the LA cohort), no significant difference was found in seclusion time for SA compared to LA antipsychotics (2.7 hours, 95% CI: -67.8, 62.5, p > 0.05). Mood stabilizer use was associated with longer seclusion (SA: 112.7 hours, LA: 215.3 hours, p < 0.05), but no difference was observed with anxiolytics.
Conclusion: Clinicians should consider individual patient needs and treatment contexts when prescribing antipsychotics. Further research is warranted to investigate broader patient outcomes and the implications of antipsychotic formulations in forensic mental healt
Enhancing Large Language Model Fine-Tuning for Classification Using Conditional Mutual Information
No full textLarge language models (LLMs) have achieved impressive advancements in recent years, showcasing their versatility and effectiveness in various tasks such as natural language understanding, generation, and translation. Despite these advancements, the full potential of information theory (IT) to further enhance the development of LLMs has yet to be fully explored. This thesis aims to bridge this gap by introducing the information-theoretic concept of Conditional Mutual Information (CMI) and applying it to the fine-tuning process of LLMs for classification tasks. We explore the promise of CMI in two primary ways: minimizing CMI to optimize a model's standalone performance and maximizing CMI to improve knowledge distillation (KD) and create more capable student models.
To implement CMI in LLM fine-tuning, we adapt the recently proposed CMI-constrained deep learning framework, initially developed for image classification tasks, with necessary modifications for LLMs. In our experiments, we focus on applying CMI to LLM fine-tuning and knowledge distillation using the GLUE benchmark, a widely used suite of classification tasks for evaluating the performance of language models. Through minimizing CMI during the fine-tuning process, we achieve superior performance on 6 out of 8 GLUE classification tasks compared to the baseline BERT model. Furthermore, we explore the use of CMI to maximize information transfer during the KD process, where a smaller "student" model is trained to mimic the behavior of a larger, more powerful "teacher" model. By maximizing the teacher's CMI, we ensure that richer semantic information is passed to the student, improving performance. Our results show that maximizing CMI during KD leads to substantial improvements in 6 out of 8 GLUE classification tasks when compared to DistilBERT, a popular distilled version of BERT
NaviX: A Native Vector Index Design for Graph DBMSs With Robust Predicate-Agnostic Search Performance
No full textThere is an increasing demand for extending existing DBMSs with vector indices to become unified systems that can support modern predictive applications, which require joint querying of vector embeddings and structured properties and connections of objects. We present NaviX, a Native vector indeX for graph DBMSs (GDBMSs) that has two main design goals. First, we aim to implement a disk-based vector index that leverages the core storage and query processing capabilities of the underlying GDBMS. To this end, NaviX is a hiearchical navigable small world (HNSW) index, which is itself a graph-based structure. Second, we aim to evaluate predicate-agnostic vector search queries, where the k nearest neighbors (kNNs) of a query vector vQ is searched across an arbitrary subset S of vectors that is specified by an ad-hoc selection sub-query QS. We adopt a prefiltering based approach that evaluates QS first and passes the full information about S to the kNN search operator. We study how to design a pre-filtering-based search algorithm that is robust under different selectivities as well as correlations of S with vQ. We propose an adaptive algorithm that utilizes local selectivity of each vector in the HNSW graph to pick a suitable heuristic at each iteration of the kNN search algorithm. We demonstrate NaviX’s robustness and efficiency through extensive experiments against both existing prefiltering- and postfiltering-based baselines that include specialized vector databases as well as DBMSs
Transformer-based Point Cloud Processing and Analysis for LiDAR Remote Sensing
No full textThe processing and analysis of Light Detection and Ranging (LiDAR) point cloud data, a fundamental task in Three-Dimensional (3D) computer vision, is essential for a wide range of remote sensing applications. However, the disorder, sparsity, and uneven spatial distribution of LiDAR point clouds pose significant challenges to effective and efficient processing. In recent years, Transformers have demonstrated notable advantages over traditional deep learning methods in computer vision, yet designing Transformer-based frameworks tailored to point clouds remains an underexplored topic. This thesis investigates the potential of Transformer models for accurate and efficient LiDAR point cloud processing.
Firstly, a 3D Global-Local (GLocal) Transformer Network (3DGTN) is introduced to capture both local and global context, thereby enhancing model accuracy for LiDAR data. This design not only ensures a comprehensive understanding of point cloud characteristics but also establishes a foundation for subsequent efficient Transformer frameworks.
Secondly, a fast point Transformer network with Dynamic Token Aggregation (DTA-Former) is proposed to improve model speed. By optimizing point sampling, grouping, and reconstruction, DTA-Former substantially reduces the time complexity of 3DGTN while retaining its strong accuracy.
Finally, to further reduce time and space complexity, a 3D Learnable Supertoken Transformer (3DLST) is presented. Building on DTA-Former, 3DLST employs a novel supertoken clustering strategy that lowers computational overhead and memory consumption, achieving state-of-the-art performance across multi-source LiDAR point cloud tasks in terms of both accuracy and efficiency.
These Transformer-based frameworks contribute to more robust and scalable LiDAR point cloud processing solutions, supporting diverse remote sensing applications such as urban planning, environmental monitoring, and autonomous navigation. By enabling efficient yet high-accuracy analysis of large-scale 3D data, this work fosters further research and innovation in LiDAR remote sensing
Numerical-Based Thermal Analysis of Proton Exchange Membrane Fuel Cells
No full textWith the urgent global concerns of climate change, governments, industry and researchers all now place a top priority on the shift to more sustainable energy technology. Fuel cells have emerged as an alternative to conventional sources of power, due to their ability to generate electricity with high efficiency and low emissions. Fuel cells are electrochemical devices that directly convert the chemical energy from fuels like hydrogen into electrical energy with high conversion efficiencies. Among various fuel cells, proton exchange membrane fuel cells (PEMFCs) have gained popularity and one of the important aspects of PEMFC is to maintain ideal operating conditions, particularly temperature for its effective operation. A numerical simulation has been conducted to analyze the temperature distribution within rectangular cooling plates with active area of 4437mm2. The performance of channels with varying width of channels, rib and header is evaluated in terms of temperature uniformity index, temperature difference and pressure drop for 15 different cases. The results show that the geometrical variations of the cooling plate play a crucial role in maintaining a uniform temperature. It was observed that the temperature uniformity index and the temperature difference decreased by approximately 30% for a higher channel-to-rib (CR) ratio compared to the base case with a smaller CR ratio. Additionally, the pressure drop for a higher CR ratio was lower than that of the other cases. To further improve the temperature distribution, design modifications were made by introducing pins at various locations in the header's inlet and outlet for 9 different cases. Although the impact of the pins was minimal due to the relatively small header size, the maximum temperature recorded was still lower than that of the baseline cases. Consequently, improved cooling performance can be effectively achieved by varying the channel geometry across the plate in parallel straight channels and through further design adjustments, such as strategically incorporating pins
Citric Acid as a Facilitator for the Integration of Okara in Soymilk Gels
No full textThis study aims to explore the effects of citric acid (CA) in the development of soymilk okara (MO) gels with improved integration of okara in soymilk protein matrix. A two-step approach was adopted. The first step focused on changes in the preparation of ingredients, soybeans, soymilk (M) and okara (O), and their effects on gel texture. The addition of CA at 0.14%, 1.25% and 2.5% (w/w) to M induced protein coagulation, significantly enhancing the mechanical and viscoelastic properties of thermally prepared soymilk gels. The particle size profile of O was manipulated by drying duration producing uniform MO gels with shorter drying time. The surface morphology of CA-treated okara (OCA), visualized with SEM, indicated that the structure of O loosened in OCA0.14 and degraded in OCA2.5.
The second step focused on the preparation of soymilk okara gels with CA at two concentrations, 0.14% and 2.5% (w/w). The role of CA on the structure of okara and the coagulation of soymilk was examined by comparing three different formulations: (i) CA treatment of okara prior to its addition to soymilk (M+OCA), (ii) CA treatment of soymilk prior to okara addition (M+CA+O), and (iii) CA addition to a mixture of soymilk and okara (MO+CA). The CA soymilk okara gels were obtained by heat treatment and their physicochemical and texture attributes were examined. The concentration and order of CA addition had different effects on the gels. FTIR, TGA and SEM analyses of CA-treated O (OCA) revealed structural modification in OCA2.5, including decomposition and release of pectic substances, which were not detected in OCA0.14.
The CA0.14 soymilk okara gels exhibited similar failure stress, failure strain and Young’s modulus, while viscoelasticity was influenced by the sequence of CA addition, in comparison to MO. Elasticity (Go’) of M+OCA and MO+CA increased threefold and viscosity (Go”) doubled, whereas M+CA+O exhibited a 4.5-fold increase in Go’ and a 3.5-fold increase in Go”. At CA2.5, the failure stress of the CA soymilk okara gels was similar to MO but with a lower failure strain and higher Young’s Modulus which indicated less cohesive, stiffer gels. The Go’ and Go” of M+OCA increased threefold and twofold, while nearly 4.5-fold in Go’ and 4-fold in Go” increases were observed for M+CA+O and MO+CA. The gel microstructure visualized with CLSM revealed changes in the protein network from strand-like in MO to particulate in the presence of CA, with CA2.5 soymilk okara gels having a denser protein network than CA0.14 soymilk okara gels. However, CLSM provided limited insights on the effect of the sequence of CA addition on gel microstructure.
In conclusion, the properties of soymilk okara gels explored in this study indicate that the gel texture can be modulated by altering CA concentration or CA sequence of addition. The CA-modification of O was not essential as similar or superior improvement in texture were achieved by pre-aggregating soymilk proteins with CA in the M+CA+O and MO+CA gel formulations
Development of Novel Surface Finishing Processes for Additively Manufactured Metal Parts
No full textPoor surface quality is one of the drawbacks of metal parts made by various additive manufacturing (AM) processes. They normally possess high surface roughness and different types of surface irregularities. Post-processing operations are needed to reduce the surface roughness to have ready-to-use parts. Among all the surface treatment techniques, electrochemical surface finishing methods have the highest finishing efficiency.
However, there are challenges with electropolishing in terms of reducing surface roughness of metals parts made via AM. Firstly, parts made with AM have both small-scale surface roughness and large-scale surface waviness. Electropolishing is only suitable for the reduction of micro-scale surface roughness while it is difficult to use the method to remove meso- to macro-scale surface waviness. In addition, it is still challenging to use electropolishing to reduce the surface roughness of internal channels of additively manufactured parts, benefiting from the promising feature of AM to produce parts with complex internal geometries. Finally, how to improve process sustainability is another question that needs to be addressed, since hazardous and corrosive chemicals are always used for the technique.
To address the aforementioned problems, novel approaches were developed, incorporating both modeling and experimental investigations. Analytical and numerical models were constructed to explore the mechanisms of electropolishing and to understand the surface evolution during the process. The results offer valuable insights that can guide the design of experiments and foster the development of novel processes.
The first experimental study focuses on using hybrid surface finishing technique to reduce meso-/macro- surface waviness. A novel surface finishing technique combining electrochemical polishing, ultrasonic cavitation and abrasive finishing was designed. Experiments were conducted on both electropolishing and hybrid finishing and the results were compared. While similar optimal arithmetic mean height values (Sa ≈ 1 μm) are achieved for both processes, the arithmetic mean waviness values (Wa) obtained from hybrid finishing are much less than those from sole electropolishing after the same processing time.
The second experimental investigation aims at electropolishing internal channels. For doing this, a novel cathode tool was invented and fabricated using polymer 3D printing. Electropolishing was conducted on both straight and curved channels with different curvatures. Preliminary experiments demonstrated a maximum surface roughness Sa reduction, from 10.86 ± 0.50 μm to 1.44 ± 0.46 μm. Apart from this, electropolishing failure mechanisms were explained and design optimization was conducted through numerical simulation. The investigations show that the method is promising in reducing surface roughness of internal channels.
In addition, experimental trials were also conducted to improve the sustainability of the surface finishing processes, including using greener electrolytes for electropolishing, and developing shear thickening polishing. Both alcohol-salt electrolyte system and deep eutectic solvent electrolyte were investigated, demonstrating effective surface roughness reduction. Shear thickening polishing using the corn starch slurry was also explored. In spite of some promising results, the process was not repeatable due to numerous influencing factors
Learning Agent-based Model Predictive Controllers for Holistic Vehicle Control
No full textHolistic vehicle control (HVC) is an advanced, integrated approach that optimizes a vehicle’s mobility, stability, and safety by coordinating all available control systems. As the automotive industry progresses towards electrification and increasing intelligence, functional integration has emerged as a dominant trend in vehicle control systems. This evolution necessitates the simultaneous coordination of multiple controllers to achieve diverse objectives. The growing demand for flexibility and reliability in automotive systems has given rise to a “plug-and-play” paradigm in control system design. This approach, while beneficial, poses significant challenges for traditional “all-in-one” integrated control methods, such as integrated model predictive control (MPC).
Distributed control schemes have demonstrated greater scalability and robustness compared to integrated schemes, especially in applications such as vehicle control systems, where managing complex, multi-agent dynamics is important. A recently proposed prominent approach within this framework is agent-based model predictive control (AMPC), where controllers are treated as interactive agents and are coordinated together to achieve a commonobjective iteratively, taking advantage of the distributed control structure. However, in practice, two critical pain points arise:
a) Uncertain contributions from unknown controllers: The optimal control performance from the AMPC highly depends on the prediction accuracy, which requires all agents or their contributions to be accurately known. This requirement is often too idealistic for practical implementation. For example, when a third party develops a “black-box” controller, its underlying algorithm is unknown, and there is no specified interface to know its contribution to vehicle dynamics. This lack of information is likely to cause a significant error in predicting vehicle behaviour, leading to unexpected or even harmful control results.
b) Limitations on controller-oriented decomposition: Decomposing from the objective’s perspective is a more practical and ideal approach in the development of function-oriented or feature-oriented automotive control systems. However, AMPC cannot decompose coupled objectives with shared agents because it is designed only to decompose the integrated system from the controller’s perspective, not from the objectives. The objective in AMPC is usually implemented as a weighted sum of multiple goals, which greatly limits the flexibility of control system design.
This thesis is hence motivated to overcome these two pain points through practical solutions using data-driven machine-learning techniques and flexible distributed schemes for multi-agent-multi-objective (MAMO) control systems.
For the first pain point, this thesis proposes a practical hybrid control scheme: learning agent-based MPC (L-AMPC). This scheme combines the model-based AMPC approach with data-driven learning methods to improve the control performance for multi-agent systems. The Gaussian process regression (GPR) enhanced by an online data management strategy serves as the learning core to predict unknown contributions along the prediction horizon, completing the system model in the MPC for more accurate control. Meanwhile, a stochastic framework is formulated to guarantee control safety and feasibility using soft chance constraints based on the prediction variance. The proposed hybrid control scheme is efficient for real-time implementation and is flexible to any control agent topology.
For the second pain point, this thesis proposes a distributed control scheme: multiobjective AMPC (MO-AMPC). This scheme adapts the alternating direction method of multipliers (ADMM) into a general control strategy that achieves global optimization while decoupling objectives. Three formulations that can maintain convergence while addressing control regularization and inequality constraints are systematically developed. The convergence and computational efficiency of the proposed methods are verified and compared on two vehicle control scenarios with multi-objective configurations.
Furthermore, this thesis proposes a data-driven distributed control scheme based on the MO-AMPC: learning multi-objective AMPC (L-MO-AMPC). To accelerate the converging process of MO-AMPC, a learning-based initialization method for iterations is proposed. The proposed scheme is compared with the MO-AMPC scheme through a path-tracking simulation using various controllers. The results show that the L-MO-AMPC scheme achieves similar control performance while significantly reducing computational time.
All proposed controllers (L-AMPC, MO-AMPC and L-MO-AMPC) are verified by real-time simulations respectively. In addition, the effectiveness and performance of L-AMPC and MO-AMPC are also verified in real vehicle experiments. The results from simulations and experiments could be concluded as follows:
• Compared to AMPC, the proposed L-AMPC can achieve higher tracking performance in well-learned scenarios with the learning capability and always guarantee constraint satisfaction even in less-learned scenarios.
• To decompose the MAMO system from the objective’s perspective, the proposed MO-AMPC achieves the same global optimum as the corresponding integrated MPC with greater flexibility and can potentially reduce the computational cost.
• Compared to model-based MO-AMPC, the proposed L-MO-AMPC has significantly higher computational efficiency while still retaining the property of converging to the global optimum, making it well-suited for real-time implementation
On Spoken Confidence: Characteristics of Explicit Metacognition in Reasoning
No full textIn this thesis, I assess how explicit, subjective evaluations of confidence influence monitoring and control (i.e., metacognitive) processes in reasoning. Metacognitive processes play a crucial role in modern dual-process theories of reasoning and decision-making, the consequences of which have been implicated in numerous significant real-world decisional outcomes. It is tacitly assumed that monitoring one’s reasoning for the purpose of optimal deployment of controlled, deliberative processing functions similarly to monitoring one’s reasoning for the purpose of providing a judgment of confidence, despite evidence from other domains indicating otherwise. This thesis takes a critical step toward evaluating metacognitive theories of reasoning and their broader application by assessing the degree to which standard approaches represent realistic accounts of metacognitive processes. To aid in interpretation of the work directly testing this possibility, I first present six experiments addressing foundational issues with respect to the operation of metacognition in reasoning. Chapter 2 provides evidence for a causal relationship between confidence judgments and controlled behavior (specifically deliberation), a reality often assumed in the absence of direct evidence. I demonstrate across four experiments that processing manipulations affect confidence and influence control behavior, consistent with a causal relationship, but also that it is possible to target control behaviour without mirroring effects on confidence. Chapter 3 develops a simple predictive model of confidence that identifies heretofore unidentified, item-based predictors of confidence. This simple model allows a unique approach to testing the central question in Chapter 4. Chapter 4 investigates whether the relationship between confidence and controlled behavior partly depends on the requirement to make explicit confidence judgments. Using a paradigm adapted from research involving nonhuman primates, I compare implicit and explicit confidence conditions. Results reveal small differences in controlled behavior and substantial differences in monitoring. In the present thesis, I provide evidence of plausibly systematic influences of common measurement approaches on reasoning. To this effect, it is likely that the reasoning processes in which individuals engage in day-to-day life are reliably different than those commonly assessed in the lab. This has practical, but also theoretical implications which I discuss
Exploring the Wellbeing and Food Security of Ethical Vegans through the Human and More-than-Human World
No full textBackground: Health and environmental data reveal significant challenges faced by populations around the word in relationship to food. Complex choices and constraints shape people’s dietary patterns. While there is often debate about which eating patterns people should follow, for vegans ethical motivations are paramount. Vegan diets can be healthy diets, but there remains a gap in the literature surrounding veganism and overall wellbeing, including food security and connection to nature.
Objectives: Taking into consideration both human and more-than-human factors this research explored how vegans live their lives embedded within the complex circumstances that shape their wellbeing. This exploration was informed by the application of a holistic model of health (the Revised Mandala of Health) and the use of an ecofeminist lens. Through a series of three studies and one methodological reflection, this research addressed the following objectives: (1) To investigate how ethical veganism influences a person’s experience and conceptualization of food security; (2) To explore the success and challenges faced by ethical vegans, how they may resist these challenges, and the implications for wellbeing; and (3) To explore ethical vegans’ experiences with the more-than-human world, and the implications for their wellbeing and coping strategies.
Methods: The first study (Chapter 2) used focus groups to learn about vegans’ experiences and ideas related to food security at the individual, household and community levels. Four focus groups were held, of which three were comprised of vegans who in the year prior had experienced food security, with the final focus group being reserved for vegans who had experienced food insecurity. The focus groups’ data were analysed using Tracy’s phronetic iterative analysis approach. The second study (Chapter 3) used semi-structured interviews to collect data on veganism and wellbeing from 26 participants. The data from these semi-structured interviews were analysed using reflexive thematic analysis (RTA). The third study (Chapter 4) had six participants who took part in both a 2-day, group, body-map storytelling (BMST) workshop and then a follow-up individual, semi-structured interview. Data collected over the course of the body-map storytelling workshop included each participant’s presentation of their body-map to the group, as well as a written ‘testimonio’, two end of day reflective exercises, and one focus group. All data from this third study were analysed together using RTA. Finally, a fourth component was generated for this dissertation (Chapter 5). This methodological reflection was written based on the lead researcher’s impressions of the BMST workshop and comments about this workshop that were shared by the participants during data collection for the third study.
Results: The results of Study 1 indicated that responsibility was a prominent feeling experienced by the participants both in relationship to their veganism and their experiences of food (in)security. The participants believed they should be informed about the food system, know how to shop frugally, and have cooking skills. Feelings of personal responsibility for food security may have been amplified by the perception of the absence of a vegan-friendly social safety net. The participants noted vegan foods could be nutritious, convenient, and inexpensive but not generally all these things at once. Therefore, compromising on at least one feature of their food was often needed. Participants in both the food secure and food insecure groups explained how at times they experienced difficulty accessing vegan-friendly foods.
The results of Study 2 demonstrated that veganism is regarded as a positive way of living that though challenging at times, especially early on in one’s experience and in relation to social relationships, was of overall net benefit to the vegan participants. This benefit contributed to enhanced wellbeing. The areas of wellbeing the participants identified most often as being influenced by veganism were the mental and emotional realms. The participants’ identities influenced their experiences of veganism. In navigating life as vegans the participants eschewed the belief in a ‘perfect’ veganism, which may have contributed to the longevity of their veganism.
The results of Study 3 showed the participants were ecologically embedded. Being ecologically embedded meant that when the participants perceived harm to the more-than-human world they subsequently felt a negative effect on their own wellbeing, including as solastalgia. However, being ecologically embedded also meant that participants experienced enhanced wellbeing through their connection to the more-than-human world.
In the methodological reflection, the researchers argue that while BMST requires significant participant involvement it can be a rewarding approach for both participants and researchers, and a safe method of data collection if careful attention is paid to the social location of potential participants and the sensitivity of the research topic. The researchers found different data collection approaches were more and less effective during the BMST workshop and they offer practical considerations for the design and undertaking of future BMST workshops.
Conclusions: Through the completion of three studies and one methodological reflection this research found that veganism was a positive force that enhanced wellbeing in the lives of the participants. As veganism is outside of the norm in Western cultures, being vegan did lead to challenges for the participants, but the challenges were not with veganism itself, rather the challenges lay in the social ramifications and accessibility of foods suitable for vegans. Participants felt the benefits of veganism outweighed the challenges. Considering the results via their alignment with the Revised Mandala of Health suggests that veganism can have health promoting properties for vegans and the more-than-human world. As plant-based diets are increasingly considered a way to address issues of health and sustainability, this study adds important understandings about the personal value and maintenance of veganism