Carleton University Institutional Repository
Not a member yet
20816 research outputs found
Sort by
Topics in the Generation of Ideals of Posets
No full textWe study the generation of the fixed ideals of finite partial orders with particular attention in finding Gray codes for their generation. Pruesse and Ruskey conjecture that the graph J(P,k), which contains as vertices the k-ideals of the poset P, with an edge between vertices that differ by a swap, has a Hamiltonian path. The conjecture is true for series-parallel posets and interval orders. We prove the conjecture also holds for the fence posets, but that the conjecture is false for the 3-ideals of the crown poset with six elements. We also provide an infinite family of posets for which the conjecture does not hold. We study the Whitney numbers of fence posets to show a different but related conjecture of Pruesse and Ruskey also holds for fences and crowns with a small number of exceptions. We study Hamiltonian cycles in the Johnson graph (n,k) with certain properties about adjacent elements which we name t-full Hamiltonian cycles. We show their connection to the recently proved Middle Levels Theorem and how to construct them. We use t-full Hamiltonian cycles to show that, for k <= n-2 the graph J(Cr(2n), k) has a Hamiltonian cycle. We also apply 2-full Hamiltonian cycles to show that (P,3) has a Hamiltonian path for any height two poset P. Further, we introduce t-full Hamiltonian connected paths, and show that 1-full Hamiltonian connected paths exist in J(n,k). We introduce a weaker version of t-full Hamiltonian paths called pair-adjacent paths, and give algorithms for their construction. We show how these algorithms can generate the 3-ideals for crown posets with more than 6 elements. Finally, we study two applications of the generation of fixed-sized ideals. We formulate the information set decoding method in terms of error-correcting codes with a poset metric, and show how the probability of success in the guessing phase of such algorithms is minimized for anti-chain posets. We generalize ordered covering arrays (OCA) for general posets, and give elementary constructions of OCAs for level-regular rooted tree posets
Learning with Limited Datasets: From Deep-Learning to Traditional Machine-Learning
Full text linkLearning with datasets containing a limited number of exemplars is a contentious research area. Researchers have used deep learning (DL) models with a large number of trainable parameters for such limited datasets leading to problems such as overfitting, over-parameterization, lack of generalization, and the need for large computational resources. Judicious use of appropriate learning methodologies may be in order when the dataset for training is limited. Non-DL methodologies or traditional machine learning methodologies with appropriate pre-processing and feature extraction techniques may perform at par or better than DL techniques for applications that have a limited dataset. This dissertation aims to establish this proposition for two healthcare applications namely, radar-based monitoring of human activities (and fall event detection) and thermography-based breast abnormality detection by developing computationally inexpensive novel supervised and unsupervised non-DL learning methodologies for binary and multi-class classification problems that outperform the current state-of-the-art techniques of the respective fields in those healthcare applications. The developed learning methodologies use traditional machine learning classifiers along with interpretable hand-crafted features such as histograms of oriented gradient (HOG), statistical features, and textural features. These novel learning methodologies use ensemble learning approaches such as early fusion, intermediate fusion, decision fusion, or training error correction. Novel contrast enhancement and novel gradient enhancement methodologies using binary encodings such as census transform and local binary patterns are also proposed to improve classification performance. For both applications, learning in the compressed domain using deterministic compressive sensing is introduced to reduce the number of trainable parameters of the developed novel supervised non-DL methodologies. The novel supervised learning methodologies using hand-crafted features achieved an average accuracy of 98% and 96% for fall event detection and breast abnormality detection, respectively. The novel unsupervised learning methodologies using hand-crafted features achieved an average error rate of 1.1% for fall event detection and an average F1-score of 85% for breast abnormality detection. At 0.875 compression ratio, the novel supervised learning methodologies in the compressed domain achieved an average accuracy of 97% and 87% for fall event detection and breast abnormality detection, respectively
Artificial Intelligence in Border Management Devices: A Multiple Correspondence Analysis of European Union Funding provided through the Horizon 2020 Program
Full text linkThis thesis answers three questions: 1) What were the most common features of devices developed during the European Union’s Horizon 2020 (H2020) research framework?; 2) Can actors/industries be associated with specific project features?; and 3) Given the emerging use of Artificial Intelligence (AI) in relation to border security devices, is the use of AI (and/or the use of certain types of AI) associated with certain features, and if so, which ones? To answer these questions this thesis uses Multiple Correspondence Analysis to analyze 42 H2020 projects which produced a border security device. These results of this show, in agreement with other literature, that projects largely conform to three clusters, those which: 1) observe territory, 2) control the flow of goods/people, and 3) protect infrastructure. Moreover, it shows that certain industries/actors can be associated with specific features and the use of AI is attributed most to projects in the first cluster
A 9dB Back-off GaN Doherty Amplifier
Full text linkThis thesis presents the design and implementation of a Doherty power amplifier operated in the Sub-7 GHz band for the purpose of improving backed off power added efficiency when compared to a single-transistor power amplifier design. The Doherty power amplifier utilizes a Wilkinson divider for an uneven input power split and a quarter-wavelength impedance inverter as the output combiner made up of lumped passive elements. The amplifier design is implemented in a underdeveloped GaN-on-Silicon technology which is modelled to have a power density of 5 W/mm. The simulated version achieves a backed-off power added efficiency of 50.5%, an improvement of over 25% compared to that of a similar sized Class-B power amplifier; and a peak saturated output power of 44 dBm is delivered to the 50-Ohm load, producing a second power added efficiency peak of 56.3%, just 1.7 dB below the maximum output power level
Understanding the Experiences and Impact in Intimate Partners of Psychopathic Individuals
Full text linkPsychopathic individuals comprise approximately 1% of the population and display maladaptive personality traits that have significant negative impacts on society. Intimate partner violence (IPV) perpetrated by individuals with psychopathic traits has been documented to be severe, versatile, and result in substantial harm to survivors. This study builds on previous findings that examine key constructs such as harms due to abuse, distress, coping, social support and posttraumatic growth. In Study 1, a scale to measure financial harm (Financial Harm Inventory; FHI) was developed and psychometrically validated in a mixed sample of current and former partners of psychopathic individuals, as well as those with friends, family or acquaintances of individuals with psychopathic traits (n = 827). The finalized 12-item scale demonstrated a unidimensional structure, with good reliability (α =.93) and excellent ability to discriminate across levels of the latent trait. The FHI demonstrated good convergent validity with measures of psychological harm (r = .32 - .35), while discriminant validity demonstrated mixed results. Study 2 surveyed respondents (n = 542) who self-identified as having previously been in a relationship with an intimate partner who has psychopathic traits. Participants reported high levels of emotional, sexual, financial, and physical harm in their relationship as well as high levels of depression, trauma symptoms and anxiety. Relationships between psychopathy, distress, coping, social support were structurally modelled. The majority of the hypothesized relationships were supported. Unexpected findings suggest ways that coping may impact distress and experiences of posttraumatic growth. Study 3 sought to better understand the experiences of survivors in a qualitative study using the Enhanced Critical Incident Technique (ECIT). Interviews (n = 15) were conducted with individuals who self-identified as being in a romantic relationship with a partner who had psychopathic traits. Themes included harmful aspects of the relationship (16 themes) and helpful or positive aspects of the relationship (11 themes). Participants also shared what resources wished they had available to them during their abusive relationship (4 themes). Themes supported previous findings of extensive harm due to the psychopathic traits of the partners and uncovered aspects of intimate partner relationships that have not been previously examined
Exploring Emotional Attachment and Engagement for the Design of Hearing Aids
Full text linkFor many people, the use of assistive technologies enhances their quality of life by augmenting their diminishing capabilities. However, some challenges with the design features cause users to reject or abandon their devices. Indeed, little attention is given to the psychological and social needs of users in the design of hearing aids, which are primarily developed according to a function-oriented medical model of disability. While the design of hearing aids has improved through technological advances such as miniaturization, there is limited design research about the emotional needs and meaningful experiences of hard-of-hearing people regarding their devices, which may lead to product abandonment. This research applies a human-centered design approach to explore users’ attitudes and experiences toward visible and expressive hearing aid designs. It seeks to understand how to create emotionally engaging and meaningful devices through design research methods. It applies an exploratory research design strategy including four studies that successively build on one another. This research contributes step-by-step tools/templates for gathering user's explicit knowledge, tacit ideas and latent needs leading to some design recommendations, criteria, and insights for the design of hearing aids. The findings suggest the importance of physical functional and aesthetic aspects of hearing aids in satisfying users’ social and psychological needs. Participants feel that a stigma is associated with the appearance of their hearing devices, and many do not find their current devices stylish, attractive, or unique. Expressive visual features could reduce stigma and enable those who like visibility to openly express themselves. Most participants have positive attitudes toward the examples of expressive visual-changing features in hearing aids. This dissertation offers the following contributions: a framework for long-term emotional attachment with assistive wearables, a human-centered design research framework with a set of tools/templates for collecting user data, an approach to data analysis, insights into the hard-of-hearing population, and a set of criteria for designing hearing aids for them
Confidence Game: How the Rules of the Confidence Relationship Impact Accountability and Executive- Legislative Relations in Parliamentary Democracies
Full text linkThe confidence connection linking cabinets to parliaments is the defining feature of parliamentary democracies. Citizens elect a parliament, parliament delegates to cabinet through an investiture vote, and cabinet retains office so long as it holds the confidence of parliament. If confidence is lost through a vote of confidence or non-confidence, the cabinet is expected to resign, opening the door to a new government or an early dissolution and election. Yet, evidence from a sample of 28 established parliamentary democracies suggests parliament has a limited capacity to carry out these functions. This is driven in part by growing democratic sensibilities and the electoral theory of democracy, that demand citizens ‘select’ the government, giving it a democratic mandate to govern, and hold it accountable at the next election. Existing research into the confidence connection is both limited and compartmentalized. It is limited because events that pull back the curtain on the confidence connection are infrequent – especially non-confidence votes with a reasonable chance of defeating the government. Compartmentalization results from the focus in the comparative literature on the individual mechanisms that parliaments use to delegate to cabinet and hold it accountable. Consequently, comparative studies isolate investiture votes, confidence votes, non-confidence votes, and dissolution rules from each other, overlooking how the rules for each mechanism were designed to work together. This project develops the confidence relationship to unite the four delegation and accountability mechanisms, plus the relevant constitutional and procedural rules, that establish the rules of the game between executives and legislatures. It then operationalizes the confidence relationship as the Parliamentary Selectoral Power Index (PSPI) to capture the relative balance of selectoral power over the confidence relationship. I find that parliamentary democracies have, despite the wide variety of rules, adopted confidence relationships that generally privilege the executive relative to parliament and cluster around the mean PSPI score. This becomes more pronounced as countries have turned towards more constructive rules that, in theory, give more power to parliament over the executive, but in practice make it more difficult for parliament to act and ultimately shift power away from parliament to a direct relationship between the executive and electorate
Adult Sperm Whale (Physeter macrocephalus) Escorts in the Eastern Caribbean Modify their Coda Vocalizations in the Presence of a Calf Compared to When with Another Adult: An Investigation of Calf-Directed Communication
Full text linkInvestigating whether adult female sperm whales use calf-directed communication by modifying their coda vocalizations in the presence of calves compared to other adults. This study analyzes coda length, coda type usage, and duration across escort-calf (EC) and escort-adult (EA) social contexts. Escorts with calves displayed greater diversity in coda type usage than when with another adult, potentially offering a broader exposure to their social unit’s vocal repertoire. Sperm whale escorts also adjusted their vocalizations by increasing coda duration when in the presence of a calf (EC) versus when with another adult (EA), likely emphasizing acoustic features to aid calf recognition and learning. This research provides evidence supporting the hypothesis that sperm whale escorts use calf-directed communication to assist calves vocal learning. These findings underscore the importance of escorts communication in calf vocal development and highlight the parallels between human language acquisition and vocal learning in non-human species
Large Language Models for the Prediction of Academic Accommodations for Students in Post Secondary Education
Full text linkEnhancing academic inclusivity for students with disabilities in post-secondary education is increasingly critical due to rising student diversity and nature of resource-intensive support systems. This study investigates the use of Language Models to analyze functional limitations and efficiently predict AAs. A chatbot interface autonomously collected qualitative and quantitative data via WHODAS 2.0 questionnaires and follow-up questions, generating a rich dataset of 170 student profiles. ML models, including transformers and LLMs were applied to recommend AAs based on these profiles, emphasizing explainability through interpretable features and LLM-generated explanations. Experiments highlighted the importance of textual data in capturing subtle functional limitations, akin to human assessments. Challenges such as data imbalance, annotation inconsistencies, and biases in recommending high-cost accommodations were identified and addressed by integrating LLM-generated explanations. This work contributes to AI and disability services by showcasing scalable, explainable, and data-driven approaches to recommending AAs
Work-Family Decision Making in Dual-Income Couples: It's all About Context
Full text linkThis thesis presents three qualitative papers that investigate decision-making experiences of mixed gender dual-income couples making challenging decisions at the work-family interface while working in a male dominated work environment (the Canadian military) with a strong organizational culture. The first paper examines the decision-making processes used as these couples made challenging decisions at the work-family interface. Social role theory and identity theory frame the analysis, This study found the organization’s ‘hypermasculine work environment’ and promotion of ideal worker norms influenced partners of both genders who want to progress at work to put work first and family second. The second study in this thesis explores how dual-income couples make the decision for one partner in the couple to voluntarily leave the case organization (the military), a profession viewed by many as a ‘calling.’ Identity theory, literature on voluntary turnover and work as a calling help interpret our findings. This study demonstrated that: the voluntary turnover decision was made by the couple, not the individual; voluntary turnover is triggered by ‘shock events’ and/or prolonged periods of job dissatisfaction; and, individuals will voluntarily walk away from a job that they consider core to their personal identity (a calling) when it conflicts with family needs, even if this decision is emotive on part of the leaver. The third study in this thesis was undertaken to help us better understand why the women in our sample were more likely than the men to decide to leave the organization (a key finding from Paper 2). This paper explored how organizational context (i.e., working within a male dominated organization) impacted how the men and women in our sample experienced gender. Findings from this study illustrate how organizational context can impact identity, and by extrapolation, couple decision-making. Social role theory and theories of ‘doing gender’ frame our understanding of the findings. Key findings were that women in the study reacted to male organizational gender norms by either doing gender/undoing military or by undoing gender/doing military. Men in the study largely denied the existence of norms influenced by the male dominated organization