Browsing by Author "Tan, Benjamin"

Now showing 1 - 9 of 9
  • Thumbnail Image
    ItemOpen Access
    A voxel-level approach to brain age prediction: A quantitative method to assess regional brain aging
    (2023-12-05) Gianchandani, Neha; Souza, Roberto; MacDonald, Ethan; Bayat, Sayeh; Pike, Bruce; Harris, Ashley; Tan, Benjamin
    Global brain age has been used as an effective biomarker to study the correlation between brain aging and neurological disorders. However, brain aging is a regional phenomenon, a facet that remains relatively under-explored within the realm of brain age prediction research using machine learning methods. Voxel-level predictions can provide localized brain age estimates that can provide granular insights into the regional aging processes. This is essential to understand the differences in aging trajectories in healthy versus diseased subjects. In this work, a deep learning- based multitask model is proposed for voxel-level brain age prediction. The proposed model outperforms the model existing in the literature and yields valuable clinical insights when applied to both healthy and diseased populations. Most findings from the analysis align with existing studies on aging, whereas other findings are intriguing and could be potential biomarkers of early-stage neurodegeneration detection. Regional analysis is performed on the voxel-level brain age predictions to understand aging trajectories of known anatomical regions in the brain and show that there exist disparities in regional aging trajectories of healthy subjects compared to ones with underlying neurological disorders such as dementia and more specifically, Alzheimer’s disease. A comparative analysis with traditional deep learning interpretability methods showed that the proposed voxel-level approach to brain age prediction is an effective way to understand regional aging trajectories while being quantitative in nature. The source code is publicly available at https://github.com/nehagianchandani/Voxel-level-brain-age-prediction.
  • Thumbnail Image
    ItemOpen Access
    Automated Bug Severity Prediction using Source Code Metrics, Static Analysis, and Code Representation
    (2022-09-12) Mashhadi, Ehsan; Hemmati, Hadi; Barcomb, Ann; Tan, Benjamin
    In the past couple of decades, significant research efforts are devoted to the prediction of software bugs. However, most existing work in this domain treats all bugs the same, which is not the case in practice. It is important for a defect prediction method to estimate the severity of the identified bugs so that the higher severity ones get immediate attention. In this thesis, we provide a quantitative and qualitative study on two popular datasets (Defects4J and Bugs.jar), using 10 common source code metrics, and also two popular static analysis tools (SpotBugs and Infer) for analyzing their capability in predicting defects and their severity. We studied 3,358 buggy methods with different severity labels from 19 Java open-source projects. Results show that although code metrics are powerful in predicting buggy code, they cannot estimate the severity level of the bugs. In addition, we observed that static analysis tools have weak performance in both predicting bugs (F1 score range of 3.1%-7.1%) and their severity label (F1 score under 2%). We also manually studied the characteristics of the severe bugs to identify possible reasons behind the weak performance of code metrics and static analysis tools. Also, our categorization shows that Security bugs have high severity in most cases while Edge/Boundary faults have low severity. Furthermore, we show that code metrics and static analysis methods can be complementary in terms of estimating bug severity. For finding the effectiveness of machine learning models in predicting bug severity, we train 8 different models on code metrics only as a baseline and evaluate them based on different evaluation metrics. The overall result was not promising, but the Decision Tree and Random Forest models have better results. Then, we leveraged the pre-trained CodeBERT model to use code representation by feeding the source code input only, and the results improved significantly in the range of 29%-140% for different metrics. We also integrated code metrics into the CodeBERT model by providing two architectures named ConcatInline and ConcatCLS which enhance the CodeBERT model efficacy.
  • Thumbnail Image
    ItemOpen Access
    Bloom Swizzlers for Efficient Keyword-based Private Information Retrieval
    (2024-10-15) Pandya, Anisha Manohar; Henry, Ryan; Reardon, Joel; Tan, Benjamin
    Private Information Retrieval (PIR) is a cryptographic primitive that allows clients to fetch information from remote databases without revealing anything about which information they are fetching to the database operators. While PIR can solve numerous privacy problems arising in our increasingly electronic society, it is a decidedly heavyweight tool whose prohibitively high latency and low throughput, coupled with its lack of expressiveness, makes it challenging to apply in real-world applications. Consequently, despite significant attention from the research community, in-the-wild deployments of PIR to date have been few and far between. This dissertation takes an important step toward making PIR more agile by introducing a novel data structure called the Bloom Swizzler. In the three decades since Chor, Goldreich, Kushilevitz, and Sudan's seminal paper on PIR (FOCS 1995), dozens of PIR constructions, proven secure under nearly as many cryptographic assumptions, have appeared in the literature. Despite the sheer number of constructions, almost all known PIR designs share a common trait: clients query for items by specifying those items' physical offsets within the database. Such a tight coupling between query construction and the underlying database's physical layout makes such PIR cumbersome for privacy practitioners and developers to use safely and correctly. Bloom Swizzlers address this by acting upon (i.e., "swizzling") clients' queries before passing them to some underlying PIR protocol, endowing almost any of the most performant PIR schemes from the literature with an efficient and intuitive key-value store interface. To accomplish this, Bloom Swizzlers synthesize ideas from two seemingly disparate constructs: Bloom Filters and Swizzlers. A Bloom Filter is a compact, probabilistic data structure that permits constant-time probable-set-membership inquiries; meanwhile, Swizzlers are a class of unary operators (represented as special 0/1-matrices) that transform vectors by rearranging, duplicating, and discarding their components. By acting as a layer of indirection between the querier and the database, Bloom Swizzlers enables a large class of PIR schemes to perform expressive, single-round keyword-based look-ups - or even query for prepared statements consisting of k-way conjunctions for relational data with multi-column keys. Moreover, with some additional effort, Bloom Swizzlers enable disjunctive queries for the value associated with any from a list of keys. Wherever a query yields more than one match, the Bloom Swizzler dictates what result gets returned; for example, one Bloom Swizzler might select a canonical representative to return (using some arbitrary, server-dictated convention) while another might return the sum or count of the matching records. We also present a proof-of-concept implementation of Bloom Swizzlers in Python and report findings from experiments designed to study the performance implications of using Bloom Swizzlers atop state-of-the-art single-server and multi-server PIR schemes.
  • Thumbnail Image
    ItemOpen Access
    Enhancing Efficiency in Residential PV Systems: Novel Maximum Power Point Tracking Strategy for Reduced DC Bus Capacitance in Differential Power Processing Architecture
    (2024-07-11) Aguero Meineri, Adrian Nicolas; Galiano Zurbriggen, Ignacio; Gray, Philippe; Westwick, David; Tan, Benjamin
    Tracking efficiency, cost, and reliability are important factors when selecting PV architectures and converter topologies. PV systems require power converters to maximize power extraction, for which DC-DC converters are a common choice. Differential Power Processing (DPP) architectures can achieve higher efficiencies and lower cost by reducing the amount of power passing through these converters, while still providing Maximum Power Point Tracking (MPPT) capabilities. Single-phase grid connected PV systems, which are the most popular choice in residential applications, require a large capacitance in the DC bus to minimize the voltage ripple caused by double-line pulsating power, which has impacts on the cost and reliability of the system. This work introduces a new MPPT mode of operation for flyback converters in DPP architectures. The proposed MPPT method shows extremely fast dynamic performance and is capable of maximizing power extraction, even for extreme variations in the bus voltage. In this way, the proposed method enables a significant reduction in the DC bus capacitance, which contributes to reducing costs and facilitating the use of ceramic capacitors, while maintaining excellent tracking efficiency. The analysis incorporates comprehensive models that characterize the large-signal dynamic behaviour of ideal and non-ideal flyback converters, and it is supported by detailed mathematical procedures. The system performance behaviour and limits are validated through simulation and experimental results.
  • Thumbnail Image
    ItemOpen Access
    Feasibility of Mapping Brain Activity to the Levels of Task Complexity within Environments of Virtual Reality
    (2023-09-21) Perez Vite, Yobbahim Javier Israel; Hu, Yaoping; Fear, Elise; Tan, Benjamin
    Mapping brain activity to certain levels of task complexity is essential for creating environments of Virtual Reality (VR), which could adapt to the mental states of human users. To investigate the feasibility of such mapping, the research work of this thesis took an approach of two steps. At first, the levels of task complexity were defined according to the geometric and appearance parameters of objects that the users interacted with for executing a task. By associating the parameters to the execution of the task, this step remedied qualitative descriptions of the levels in current state-of-the-art. Secondly, an empirical study of two experiments was conducted within a VR to collect brain activities (as brainwaves) of human participants (i.e., users) during the execution involving various task complexity. Using a device of encephalography (EEG) to collect the brainwaves, this step assessed several existing features derived from the brainwaves as potential indicators of feasibility. This thesis produced two significant findings: (1) the definition of task complexity is quantitative and could be suitable for describing object-oriented tasks, and (2) specific EEG features – such as engagement ratio – could indicate increased or decreased levels of task complexity. Hence, the work indicates the feasibility of mapping brain activity to the levels of task complexity. Future investigations are needed to refine the definition, and EEG features for optimizing cognitive engagement and performance by modulating the levels of task complexity. The outcomes of the investigations could have implications for training, simulation, and user experience in various VR-based applications.
  • Thumbnail Image
    ItemOpen Access
    Sport and Recreational Activity Participation and Injury Risk in Elementary School Children with Probable Developmental Coordination Disorder or Attention Deficit Hyperactivity Disorder
    (2014-09-29) Tan, Benjamin; Emery, Carolyn; Dewey, Deborah
    Objective: To examine if probable Developmental Coordination Disorder (pDCD) or Attention Deficit Hyperactivity Disorder (pADHD) in elementary school children aged 8 – 13 years is associated with sport and recreation (S&R) participation and injury. Methods: Cross-sectional design. Recruitment from elementary schools in Calgary, Alberta, Canada led to 681 participants completing an anonymous questionnaire. Primary Outcome Measure: S&R participation and injury. Results: S&R participation was reported by 82.7% (95%CI; 79.8, 85.5) of children; those with co-occurring pDCD and pADHD participated less than their typically developing peers. The injury incidence rate was 2.43 injuries/1000 participation hours (95%CI; 2.06, 2.85), with no differences between study groups. Ethnicity, stressful life events and coaching were potential risk factors for injury identified by exploratory risk factor analysis. Conclusions: Children with pDCD and/or pADHD were not at greater risk of S&R injury than their typically developing peers, but those with co-occurring pDCD and pADHD participated in S&R less.
  • Thumbnail Image
    ItemOpen Access
    The Canadian Registry for Pulmonary Fibrosis: Design and Rationale of a National Pulmonary Fibrosis Registry
    (2016-04-05) Ryerson, Christopher J.; Tan, Benjamin; Fell, Charlene D.; Manganas, Hélène; Shapera, Shane; Mittoo, Shikha; Sadatsafavi, Mohsen; To, Teresa; Gershon, Andrea; Fisher, Jolene H.; Johannson, Kerri A.; Hambly, Nathan; Khalil, Nasreen; Marras, Theodore K.; Morisset, Julie; Wilcox, Pearce G.; Halayko, Andrew J.; Khan, Mohammad Adil; Kolb, Martin
    Background. The relative rarity and diversity of fibrotic interstitial lung disease (ILD) have made it challenging to study these diseases in single-centre cohorts. Here we describe formation of a multicentre Canadian registry that is needed to describe the outcomes of fibrotic ILD and to enable detailed healthcare utilization analyses that will be the cornerstone for future healthcare planning. Methods. The Canadian Registry for Pulmonary Fibrosis (CARE-PF) is a prospective cohort anticipated to consist of at least 2,800 patients with fibrotic ILD. CARE-PF will be used to (1) describe the natural history of fibrotic ILD, specifically determining the incidence and outcomes of acute exacerbations of ILD subtypes and (2) determine the impact of ILD and acute exacerbations of ILD on health services use and healthcare costs in the Canadian population. Consecutive patients with fibrotic ILD will be recruited from five Canadian ILD centres over a period of five years. Patients will be followed up as clinically indicated and will complete standardized questionnaires at each clinic visit. Prespecified outcomes and health services use will be measured based on self-report and linkage to provincial health administrative databases. Conclusion. CARE-PF will be among the largest prospective multicentre ILD registries in the world, providing detailed data on the natural history of fibrotic ILD and the healthcare resources used by these patients. As the largest and most comprehensive cohort of Canadian ILD patients, CARE-PF establishes a network for future clinical research and early phase clinical trials and provides a platform for translational and basic science research.
  • Thumbnail Image
    ItemOpen Access
    Towards Reconfigurable Hardware for In-field Hardware Bug Patches
    (2024-09-16) Dharavathu, Anudeep; Tan, Benjamin; Murari, Kartikeya; Yanushkevich, Svetlana
    System-on-chip (SoC) designs are becoming increasingly complex, and the ability to detect and address all possible bugs at design time is highly challenging. Thus, to improve the survivability of SoC designs, it is desirable to be able to patch newly discovered design bugs or potential vulnerabilities in the field. Recently, the idea of hardware-based patching, especially of hardware bugs, has emerged as a complementary approach to software/firmware-based post deployment updates. In anticipating potential problems, designers must invest an upfront cost to implement hardware-based patching infrastructures. My thesis investigates the feasibility of incorporating an embedded field-programmable gate array (eFPGA) fabric as an approach to enable hardware-based patching, i.e., reprogrammable hardware to patch hardware bugs, and explores the resource overhead costs for varying patching architectures. We also characterized different eFPGA configurations to help designers decide the eFPGA design parameters.
  • Thumbnail Image
    ItemOpen Access
    Utilization of Natural Language Processing for Extracting Smart Cities Requirements from Large Social Media Text
    (2024-05-14) Mirshafiee Khoozani, Mitra Sadat; Barcomb, Ann; Tan, Benjamin; Messier, Geoffrey; Fapojuwo, Abraham
    Major organizations such as urban centers worldwide face challenges from rapid population growth and evolving demands, requiring innovative approaches to stay responsive to residents' needs. This challenge is exemplified by the city of Calgary, where an automated system for aggregating and categorizing resident feedback could improve city planning. What people find important and useful can be seen in the articles they post on social media. One method for determining the performance of urban services and assets for citizens is paying attention to these data generated by the residents. In this regard, we need to examine datasets wherein writing is the primary form of citizen engagement (direct messages, requests, comments, complaints, etc.). To interpret this data, it is necessary to use appropriate tools and techniques for data processing and analysis of large volumes of unstructured text. Some of the most effective tools used by researchers nowadays falls into the scope of computational linguistics, specifically Natural language processing (NLP). Furthermore, Twitter is one of the primary platforms where individuals freely voice their opinions and concerns. In this study, we develop an automated workflow that can scrape, classify, and display tweets in a simplistic view. With the help of this system, local officials will be able to speed up the decision-making process when considering citizens' current problems. Following our research question, we look into the optimal scraping criteria, explore a variety of methods for topic and emotions analysis, and validate these methods both using automatic evaluation and manual assessment. As a result, we are able to identify issues related to city development, senior citizens, taxes, and unemployment using our best performing models (BERTopic for topic modeling and few-shot learning using Setfit for emotion analysis.) Afterward, we collect city employees' opinion regarding our research to determine the usefulness and applicability of this approach. Overall, we demonstrate how delving into these analyses can complement the current systems in place for urban planning.