PRISM | Institutional Repository
Communities in PRISM
Select a community to browse its collections.
Recent Submissions
Federated Learning Model Aggregation in Heterogeneous Aerial and Space Networks
(2024-10-09) Dong, Fan; Drew, Steve; Leung, Henry; Drew, Steve; Leung, Henry; Ye, Qiang; Wang, Mea
Federated learning offers a promising solution for overcoming the challenges of networking and data privacy in aerial and space networks by harnessing large-scale private edge data and computing resources from drones, balloons, and satellites. Although existing research has extensively explored optimizing the learning process, improving computing efficiency, and reducing communication overhead, statistical heterogeneity remains a substantial challenge for federated learning optimization. While state-of-the-art algorithms have made progress, they often overlook diversity heterogeneity and fail to significantly improve performance in high-degree label heterogeneity conditions. In this thesis, statistical heterogeneity is further dissected into two categories: diversity heterogeneity and label heterogeneity, allowing for a more nuanced analysis. It also emphasizes the importance of addressing both diversity heterogeneity and high-degree label heterogeneity in aerial and space network applications. A theoretical analysis is provided to guide optimization in these two challenging scenarios. To tackle diversity heterogeneity, the WeiAvgCS algorithm is introduced to accelerate federated learning convergence. This algorithm employs weighted aggregation and client selection based on an estimated diversity measure, termed projection, enabling WeiAvgCS to outperform other benchmarks without compromising privacy. For high-degree label heterogeneity, the FedBalance algorithm is proposed, utilizing the label distribution information of each client. A novel metric, termed relative scarcity, is introduced to determine the aggregation weights assigned to clients. During the training process, fully homomorphic encryption is employed to protect clients’ label distributions. Additionally, two communication protocols are designed to facilitate training across different scenarios. Extensive experiments were conducted, demonstrating the effectiveness of WeiAvgCS and FedBalance in addressing the research gaps in diversity heterogeneity and high-degree label heterogeneity.
A Case of One: An Autobiographical Design Approach to Explore a Personal Informatics Preparation Stage Procedure
(2024-09-20) Zhang, Xinchi; Schroeder, Meadow; Ringland, Kathryn; Zhao, Richard; Wang, Mea
This thesis exploration was started as a personal design endeavor to have a system that can support realistic task arrangement during my graduate school. This exploration landed to the often-overlooked area — the preparation stage in the Stage-Based Model (SBM) — in the personal informatics (PI) field. Personal informatics supports people to gain self-understanding through reflection on their relevant personal data. The preparation stage, which can involve many decision-making processes such as understanding the motivation of collecting personal information, deciding the information to collect, and choosing the appropriate tools, is where prior PI research focused significantly less on. This thesis aims to narrow this gap by introducing a procedure and an accompanying artifact, Qubio. I took an autobiographical design approach. Autobiographical design offers many advantages such as close use to allow rapid iteration whenever needed (fast tinkering). Then, combining with reflection, diligent documentation (46+ hrs recordings, 262 reflection entries), and long-term usage (47 months), I established a personal reflective procedure to determine what data I might track. The procedure includes 1) externalization of obligations and interests, 2) mapping (for goal choices), and 3) task arrangement, which is supported by the token-based artifact, Qubio. This exploration bridges the preparation stage of the Stage-Based Model in PI and the Integrated Model of Goal-Focused Coaching (Integrated Model) in psychology. I conclude this thesis by discussing research opportunities in connection to the Integrated Model for the preparation stage in PI and suggesting collaboration between PI and personal information visualization to support visualization agency in PI practices. I further suggest revisiting established PI models to potentially integrate the field’s expanded understanding of PI related activities. Finally, I reflect on how an autobiographical design approach produced a personalized procedure and artifact.
The Metabolism of Uropathogenic Bacteria in in Vitro Human Urine Cultures
(2024-10-07) Chan, Carly C. Y.; Lewis, Ian A.; Turner, Ray J.; Harrison, Joe; Chaconas, George; Montenegro-Burke, Rafael
Urinary tract infections (UTIs) are common infections primarily caused by bacterial colonization of the host’s bladder and/or kidney. Research into the molecular underpinnings behind UTIs primarily focused on the various macromolecular virulence factors that enable uropathogens to invade and colonize the host’s urinary tract. As such, there is an extensive body of literature characterizing these UTI-associated virulence factors. However, one important aspect that remains relatively unexplored is pathogen metabolism. Pathogens must be able to metabolize the nutrients available in its microenvironment to survive and grow within their host. Human urine is a chemically complex medium with a diverse range of amino acids and nucleic acids, but generally lacks carbohydrates, the preferred carbon source for most microbes and thus, urine is often considered a nutrient-poor substance. Recent technological advancement in metabolomic tools can allow researchers to discover new insights into uropathogen metabolism and further our understanding of how uropathogens overcome nutritional adversity and survive in human urine. To address this gap, I used liquid chromatography-mass spectrometry (LC-MS) to analyze in vitro human urine cultures of uropathogenic bacteria. My initial metabolomics survey of eight common uropathogenic species found that these species can be divided into four distinct metabolic clades: serine consumers, glutamine consumers, amino acid abstainers, and amino acid minimalists. There were also several other metabolic phenotypes exclusive to a single or a few species. Metabolites found to be secreted by uropathogens could be candidate UTI biomarkers. Previous work discovered that agmatine was a robust UTI biomarker for several bacterial species in the Enterobacterales order including Escherichia coli. Investigation into bacterial agmatine production revealed that E. coli utilizes several different decarboxylase-based acid resistance systems in urine, induced at different pH ranges. Meanwhile, a few non-Enterobacterales species, like Staphylococcus spp., were found to secrete N6-methyladenine, which was also identified as a UTI biomarker. Advances to metabolomics methods can greatly enhance the efficiency of these metabolomics analyses, and so we developed a new LC-MS strategy for monitoring microbial metabolic activity in real-time. By investigating uropathogen metabolism with current metabolomics tools, we can better understand how these pathogens may persist and thrive during UTIs.
Three-Dimensional Building Reconstruction from ALS Point Clouds
(2024-10-04) Yang, Hongxin; Wang, Ruisheng; Wang, Xin; Hassan, Quazi Khalid; Yang, Hongzhou; Cheng, Yufeng(Frank); Yang, Bisheng
Reconstructing buildings from Light Detection and Ranging (LiDAR) point clouds obtained from aerial perspectives is of significant importance in the domain of photogrammetry. Given that the experimental dataset, Building3D, lacks sufficient corner points and exhibits point cloud sparsity among other challenges, point cloud completion (PCC) techniques, a branch of reconstruction, are employed to complete the building facade information. Due to the high demand for labeled data and the associated high cost of manual annotations, Self- Supervised Learning (SSL) methods for three-dimensional (3D) point clouds have garnered considerable attention from scholars. However, existing methods commonly use a standard Transformer backbone, result- ing in quadratic time complexity. To overcome these limitations, an innovative masked linear autoencoder framework is proposed. Due to the storage requirements—approximately 400:4:1 for point cloud, mesh, and wireframe formats, respectively, wireframe models have recently garnered considerable attention in the field of remote sensing. Despite some early explorations into constructing wireframe models, numerous challenges persist. This thesis revisits 3D building wireframe reconstruction from a SSL perspective, with the aim of alleviating or even addressing these existing difficulties. A two-stage Self-supervised (SS) pretraining architecture is proposed to generate wireframe models. Initially, it utilizes a SSL-based pretraining framework that incorporates a linear self-attention mechanism (SAM) to generate point-wise features. Subsequently, corner detection and edge prediction modules are employed to classify and regress the coordinates of corner points and to determine optimal edge selections, respectively. To address the issue of insufficient corner points, a SSL-based pretraining method for 3D wireframe reconstruction, guided by an edge point regression module, is proposed. The parameters of the wireframe’s edges—including edge length, direction vector, and direction offset—are regressed under the guidance of the edge point regression module. To enhance the clustering of roof wireframe vertices, an efficient approach based on a multiclass TWin Support Vector Machine (TWSVM) framework is proposed. This framework aims to simplify the model by effectively clustering roof wireframe vertices.
An Intersectional Approach to Addressing Discrimination Among Pregnant Individuals in a Contemporary Population: A Cross-Sectional Study
(2024-09-25) Dharamsi, Shazia; Metcalfe, Amy; Ahmed, Sofia; Ruzycki, Shannon
Background: Pregnant individuals may be vulnerable to discrimination due to existing social stigma and cultural norms surrounding pregnancy and motherhood, which can be exacerbated due to gender, race, or other aspects of their identity; however, it is unclear how intersecting identities overlap and interact to shape peoples’ everyday experiences. The purpose of this study was to compare the prevalence and experiences of discrimination among pregnant individuals with diverse intersectional identities. Methods: This secondary analysis used baseline data from 1,605 participants enrolled in a longitudinal pregnancy cohort study, the P3 Cohort Study. Latent class analysis was applied to identify different subgroups within the population based on combinations of their social positions and identities (i.e., race, gender expression, emotional health, physical health, income, disability, chronic illness, and age). Perceived discrimination was assessed via the Everyday Discrimination Scale (EDS), and linear regression was performed to compare the frequency, chronicity, and overall number of discriminatory encounters between subgroups. Results: Utilizing latent class analysis, the following three subgroups emerged: (1) Mostly Privileged and Healthy; (2) Somewhat Privileged with Chronic Conditions; (3) Somewhat Privileged, Predominantly Racialized. Class 2 (βfrequency = 0.8, p = 0.042; βchronicity = 45.2, p < 0.001; βsituations = 0.4, p = 0.040) and Class 3 (βfrequency = 0.2, p < 0.001; βchronicity = 31.5, p < 0.001; βsituations = 0.5, p < 0.001) were both positively associated with the frequency, chronicity, and total number of discrimination situations relative to Class 1. The primary reported reasons underlying discrimination experiences varied across latent classes, reflecting the intersecting identities that defined each cluster.