oURspace (Univ. of Regina)
Not a member yet
15455 research outputs found
Sort by
Oil and Non-Oil Determinants of Saudi Arabia’s International Competitiveness: Historical Analysis and Policy Simulations
No full textTo achieve sustainable economic growth, Saudi Vision 2030’s target is to improve Saudi Arabia’s ranking on the Global Competitiveness Index from 25 in 2015–2016 to within the top 10 by 2030. Saudi Arabia also aims to increase the share of non-oil exports in the non-oil GDP from 16% in 2016 to 50% by 2030. For policymakers to make informed decisions to achieve these goals, they need to understand the driving forces of Saudi Arabia’s competitiveness. To this end, we consider the real effective exchange rate (REER) as a measure of external price competitiveness, as it captures domestic and global price changes. We then examine the REER using a two-stage modeling framework. First, we estimate the REER equation, which allows us to assess the impacts of the determinants and evaluate currency misalignments as a competitiveness indicator. Second, we extend the KAPSARC Global Energy Macroeconometric Model (KGEMM) with the estimated equation, which provides a framework for simulating the competitiveness impacts of the theoretically formulated determinants and other variables relevant to policymakers. The framework also allows us to account for feedback loops. We conduct a policy scenario analysis to quantify the competitiveness effects of the Public Investment Fund’s (PIF) new strategy for 2021–2025. We derive the following policy insights. Authorities may wish to implement initiatives boosting future productivity and, thus, competitiveness, such as PIF investments. Policymakers should be regularly informed about currency misalignment. Government consumption and public investment projects should consider substituting imports with locally produced goods and services. Local content development would also help to diversify the Saudi economy. Finally, attracting more foreign investment and other assets from the rest of the world may lead to technological development and improvement in the economic, financial, and social infrastructure and business environment, all enhancing competitiveness.The authors are grateful for receiving open access funds from the President’s Publication Fund and Tri-Agency Cohort Fund at the University of Regina
The environmental and geochemical controls on mercury and sulphur cycling in Prairie Pothole Region wetland complexes
No full textA Thesis Submitted to the Faculty of Graduate Studies and Research In Partial Fulfillment of the Requirements for the Degree of Master of Science in Geology, University of Regina. x, 119 p.Wetlands are known to play important roles in the cycling of mercury (Hg), which may
ultimately lead to the generation of methylmercury (MeHg), a neurotoxin that bioaccumulates in
food webs. The Prairie Pothole Region (PPR) of North America hosts sulphur-rich wetlands that
favour the production of MeHg. Within these wetland systems, dissolved organic carbon (DOC)
and sulphate fuel bacterial sulphate reduction (BSR), resulting in increased Hg methylation in
select PPR wetlands. Due to local variability in wetland geochemistry, a reflection of hydrologic
processes and differences in surficial Quaternary lithology, understanding the biogeochemical
cycling of Hg in these wetlands is inherently complex. In Saskatchewan, PPR wetlands vary with
respect to surface water and sediment chemistry, groundwater interactions, topographic position,
and permanence, among other factors. As a result, MeHg can vary significantly on a wetland-bywetland
basis in even the smallest complexes. Here, the geochemical and hydrogeological controls
of sulphur cycling on Hg methylation in the St. Denis National Wildlife Area (SDNWA), an
analogue for the northern Prairie Pothole Region, were explored. This was done through evaluating
groundwater–surface water interactions, predicted Hg speciation, and how the sulphur cycle
related to Hg methylation in the SDNWA. Previously collected surface and groundwater chemistry
datasets obtained from the Global Institute of Water Security (University of Saskatchewan) and
previous studies conducted at the SDNWA were coupled with new water and bulk sediment
geochemistry analyses and sulphur stable isotope analyses. Results indicate that varying
topographic position, wetland type (recharge or discharge), and prevailing bottom water and
sediment redox conditions fundamentally influence sulphur cycling and Hg methylation in the
SDNWA. In addition to varying bottom water redox conditions, Hg-S speciation, and trace metal
cycles in the wetlands, indicate that localized variability may strongly influence overall Hg
methylation. In calcium sulphate wetlands, sulphur disproportionation and sulphide mineral
formation may keep porewaters from becoming too concentrated in sulphide and stifling BSR and,
by extension, MeHg production. In contrast, the Hg-S speciation, temporary wetland nature, and
the lower concentrations of sulphate in the surface waters of calcium bicarbonate wetlands may be
more favorable for BSR and result in transient periods of Hg methylation. Although porewater and
surface water redox characteristics played a large role in Hg methylation, other processes could
significantly influence net methylation rates and overall ecosystem health. As such, constraining
the biogeochemical cycling of Hg in PPR wetlands is critical for understanding how high levels of
MeHg may impact aquatic biota.Studentye
Publishing linked open data from spreadsheets
No full textA Thesis Submitted to the Faculty of Graduate Studies and Research In Partial Fulfillment of the Requirements for the Degree of Master of Science in Computer Science, University of Regina. x, 108 p.The World Wide Web (WWW), as defined by the World Wide Web Consortium (W3C), is “the universe of network-accessible information, the embodiment of human knowledge”1. With development over the years, the global web of linked documents has become the global web of linked data, which is popularly known as the “semantic web”.
The semantic web is now an established topic of research but there remain many unanswered questions. Data that are linked and open facilitate discovery and reuse. Likewise, vocabularies used to describe the semantic relationships amongst linked and open data facilitate discovery and reuse. Yet, it is not always a straightforward matter to take data that may be somehow accessible on a webpage written in Hyper Text Markup Language (HTML) and publish it as Linked Open Data (LOD). Many complimentary technologies are used in handling Linked Open Data (LOD) on the web, particularly the Resource Description Framework (RDF), the Web Ontology Language (OWL), and the Simple Protocol and RDF Query Language (SPARQL), to name a few.
Tim Berners-Lee’s vision for a global hypertext system continues to be fulfilled. This thesis work contributes to the practice of publishing Linked Open Data (LOD) by following the lifecycle of a particular dataset from spreadsheet to Linked Open Data (LOD) that is accessible on the web. As such, this thesis can serve as a guide for researchers and citizens-at-large who want to contribute their data to the world. It talks about the methodology that can be used to describe a linked open vocabulary to map survey data from a spreadsheet file.
1. https://www.w3.org/WWW/Studentye
Quantification of crops' consistency on corn fields using robust deep learning models
No full textA Thesis Submitted to the Faculty of Graduate Studies and Research In Partial Fulfillment of the Requirements for the Degree of Master of Applied Science in Electronic Systems Engineering, University of Regina. xiv, 114 p.Inconsistent crops across large fields lead to lower overall crop productivity. Traditional
methods for evaluating the consistency of crops, such as manual inspection of a predefined
sample area to determine plant stand count, crop coverage, and spacing, are time-consuming
and inaccurate, leading to insufficient analysis of the crops’ status and poor decision-making.
Despite the recent technological advancement in the agricultural management system, an
automated approach to assess multiple parameters has not been thoroughly explored. This
thesis aims to develop a vision-based framework to identify uneven crop distribution areas
and quantify the consistency of crops across different fields under natural conditions. It
utilizes two convolutional neural networks - the state-of-the-art object detection algorithm
- YOLOv7 to detect and precisely count corn plant stands per row combined with an
improved semantic segmentation model based on the U-Net architecture for crop row
segmentation. This improved semantic segmentation model is further integrated with
morphological operators and thinning algorithms to obtain a single-pixel representation of
the crop rows. Once the crop rows are extracted, the count and location of corn plant stands
per row are determined using YOLOv7.
Furthermore, by incorporating the Distance-IOU non-maximum suppression technique
into YOLOv7, we can suppress redundant boxes by considering both the detection box’s
overlapping area and center points. This leads to more effective detection of the target crops
even when occluded. The inter-plant distance estimation is obtained using this information,
followed by standard deviation calculation. The proposed segmentation model in this
framework is further used to determine the leaf area index, providing key insight into crop
coverage. The suggested architecture for semantic segmentation exploits the ResNet34
module into the U-Net encoder to extract features efficiently during the down-sampling
process. This results in a balance of lower trainable parameters and the preservation of
spatial resolution during the feature extraction process. Our framework demonstrates a high
mean average precision of 0.939 in identifying plant stands. Furthermore, the segmentation
models employed for obtaining crop densities and crop row detection achieve a mean
intersection over union score of 0.887 and 0.859, respectively. Our results demonstrate
the robustness of the proposed framework in diverse crop fields and its ability to adapt to
variations in crop conditions.Studentye
Understanding PTSD among correctional workers in Manitoba, Canada: Key considerations of social variables
No full textMounting evidence highlights the high prevalence of posttraumatic stress disorder (PTSD) among correctional workers. The current analysis draws on survey response data to present a social profile of correctional workers in the province of Manitoba (n = 580), Canada, who screened positive for PTSD (n = 196). We examined demographic information, professional history information, and adverse work exposure experiences, as well as treatment and support patterns. The analysis was not intended to identify correlates of PTSD development among correctional workers, but did identify the characteristics, professional and personal situations, and treatment experiences of correctional workers who screened positive for PTSD. The results highlight the multidimensional nature of work stressors, the pronounced problem of work–life conflict, and variations in seeking supports and treatments. Generally, participants screening positive for PTSD reported higher exposure to potentially psychologically traumatic events, higher environmental or occupational stressors at work, and many had prior work experience as public safety personnel. Correctional workers who screened positive for PTSD appeared more likely to access mental health supports. Promoting proactive support seeking for mental health treatment may help to mitigate the severity, frequency, stigma, and length of mental health challenges among correctional workers.Facultyye
Patient narratives in internet therapy
No full textA Thesis Submitted in Partial Fulfillment of the Requirements for the Degree of Bachelor of Science (Honours) in Psychology, University of Regina. 46 p.Background. Previous findings have shown positive treatment outcomes for Canadian public safety personnel (PSP) in the internet-delivered cognitive behavioural therapy (ICBT) program entitled the PSP Wellbeing Course. Despite this, PSP have provided mixed reviews regarding their satisfaction with the patient narratives, or Case Stories, that are dispersed throughout the course.
Objectives. In an attempt to improve treatment engagement and outcomes, the current study explored patient perceptions of Case Stories and elucidated the relationship between review of the Case Stories and outcomes. We hypothesized PSP would perceive the Case Stories as a source of education, persuasion, motivation, engagement, and comfort, aligning with themes commonly found in the literature on patient narratives.
Methods. We administered pre- and post-treatment questionnaires to 23 PSP who took the PSP Wellbeing Course and asked PSP to rate the Case Stories at post-treatment. To gather a richer understanding of PSP’s opinions of the Case Stories, we conducted 10 semi-structured interviews. We analyzed data using a mixed-methods approach.
Results. Of the 13 participants (n = 57%) who reviewed the Case Stories, most agreed that the Case Stories allowed them to feel less alone in their experiences. While clients experienced statistically significant reductions in mental health symptoms, there was no significant difference in treatment outcomes between those who reviewed the Case Stories and those who did not. . Participants who were interviewed identified the Case Stories as being relatable, a useful psychoeducational tool, and a source of comfort. Some participants provided suggestions to improve the Case Stories (e.g., additional content, audiovisual aids).
Discussion. The current study provides novel insight into the role Case Stories have in ICBT and may inform improvements to use of case stories in ICBT.Studentn
Precision-based boosting for regression
No full textA Thesis Submitted to the Faculty of Graduate Studies and Research In Partial Fulfillment of the Requirements for the Degree of Master of Science in Computer Science, University of Regina. xi, 75 p.Regression is a type of predictive modeling problem that involves estimating a continuous
numerical value based on input variables. The goal of this research is to
investigate whether incorporating the precision of regression models on specific target
values can improve the performance of ensemble-based regression models.
We begin by reviewing two existing ensemble methods for classification, namely
AdaBoost and PrAdaBoost, which will form the basis of our proposed ensemble
method for regression. We also provide a formal analysis of the training error upper
bounds for PrAdaBoost and AdaBoost. The mathematical proof shows that PrAdaBoost’s
upper bound is always less than or equal to AdaBoost’s. This result is important
because it implies that PrAdaBoost’s training error upper bound decreases
exponentially as the number of iterations increases, assuming that each individual
predictor in the ensemble is better than random guessing.
We modify the PrAdaBoost algorithm and implement it in the context of regression,
thus introducing a new regression algorithm called PrSAMME-R. To evaluate
the performance of PrSAMME-R, several experiments are conducted on various
regression datasets, and the results are compared to those obtained from other
ensemble-based regression models.
The results show that incorporating the precision of regression models on specific
target values into their weights can improve the performance of ensemble-based
regression models significantly. PrSAMME-R outperforms other ensemble-based rei
gression models such as Random Forest, Gradient-based Boosting, AdaBoost.R2 and
AdaBoost.RT, in terms of mean absolute error.Studentye
Are expectations of obstructed facial features accurate?
No full textA Thesis Submitted in Partial Fulfillment of the Requirements for the Degree of Bachelor of Arts (Honours) in Psychology, University of Regina. 22 p.To fill in missing facial information of partially obstructed, unfamiliar faces, it is believed that people form an accurate holistic expectation. In support of this claim, Winand (2022) demonstrated that participants could correctly match the bottom half of a face to its top half. Yet, the study is limited by the fact that participants may have been able to match the face halves using superficial characteristics such as shading and texture rather than the shapes and sizes of features. Therefore, the purpose of this study was to replicate Winand’s (2022) findings with a task in which such superficial matching strategies cannot be used. This was achieved by morphing images together to blur out such inconsistencies in the stimuli. Undergraduates (N=122) were shown the top and bottom halves of a face that belonged to either a single-identity (two photos of the same identity morphed together) or dual-identity (two different identities morphed together). Participants toggled between two randomly chosen bottom halves that belonged to either the same people shown in the top half or different people, and chose the bottom half that best matched the top half. Overall, accuracy was well above chance, but highest when choosing the best single-identity bottom half for a single-identity top. Thus, although incorporating another identity decreases accuracy, people are generally able to find similarities among the top and bottom half identities without the aid of superficial characteristics. This suggests that people accurately form a holistic expectation based on the available top features.Studentn
Optimizing golf ball detection by applying channel pruning to a YOLOv3-Tiny detector
No full textA Thesis Submitted to the Faculty of Graduate Studies and Research In Partial Fulfillment of the Requirements for the Degree of Master of Science in Computer Science, University of Regina. xvii, 160 p.This thesis presents an approach to golf ball detection using a channel pruned YOLOv3-
Tiny detector. Object detection is a challenging task, influenced not only by the
appearance of objects but also by the surrounding environment. In the case of golf
ball detection, the object can vary in size, color, and texture, and may be partially
or fully occluded by other objects in the scene. Additionally, environmental factors
such as lighting conditions and background clutter can further complicate the task.
In addition, the fast motion of the golf ball adds another level of difficulty to the
detection process. The proposed approach utilizes channel pruning to reduce the size
and computational complexity of the YOLOv3-Tiny detector while maintaining high
detection effectiveness.
In this thesis, we propose an approach to improve the effectiveness and efficiency
of golf ball detection using YOLOv3-Tiny. We apply channel pruning, which involves
removing unimportant channels from the network, to reduce the size of the model and
speed up inference. We begin by training YOLOv3-Tiny using sparse training, where
we consider the scaling factor for each batch normalization layer in the architecture
of the model. These scaling factors represent the importance of each channel, and we
use the L1-regularization process to keep them small during training. After training,
we prune channels with small values and evaluate the pruned model. This approach
leads to a more efficient and effective golf ball detection model.
We also present a new definition for pruning convolutional neural networks, which
considers two aspects. Firstly, it considers the maximum amount of pruning that can
be performed while maintaining a certain level of predictive effectiveness. Secondly, it
determines the maximum level of predictive effectiveness that can be achieved while
keeping a specific amount of pruning. This approach allows us to fine tune the pruning
process to balance efficiency and effectiveness.
The proposed method for golf ball detection using channel pruning in YOLOv3-
Tiny offers several advantages. Firstly, it does not require adding any extra coefficient
or parameter to the model for sparse training, as the parameters in the batch normalization
layer are used as representatives of the importance of channels. Secondly, by
incorporating two thresholds in the pruning section, the architecture of the model is
preserved and overpruning is avoided. Furthermore, the flexibility of setting these two
parameters to different values according to the specific application and desired effectiveness
and efficiency limits adds to the method’s versatility. Finally, the proposed
approach can be extended to other object detection problems, making it a promising
technique with a wide range of applications.
In our evaluation, we conducted experiments on two different platforms, GPU
and CPU, to observe the effect of pruning on both. We also tested our model on
different image sizes. Our results showed a significant improvement in the parameter
size for both platforms. In particular, we observed a remarkable improvement in
CPU inference time. Our pruned model achieved a 91% reduction in the number of
parameters, which significantly reduced the memory usage of the model. Moreover,
our pruned model outperformed the unpruned model trained with sparse training
by 1.7% in terms of F1 score. These results demonstrate the effectiveness of our
proposed approach in improving the effectiveness and efficiency of golf ball detection
using YOLOv3-Tiny.Studentye
Frequency domain analysis of U-Net segmented ultrasound images
No full textA Thesis Submitted to the Faculty of Graduate Studies and Research In Partial Fulfillment of the Requirements for the Degree of Master of Science in Computer Science, University of Regina. x, 71 p.During prostate cancer brachytherapy, catheters are inserted into a patient's
prostate for a highly localized radiation treatment. Accurately placed catheters are
critical for successful treatment and ultrasound images are taken throughout the procedure
to verify their exact positions. However, manually locating catheters on ultrasound
images is extremely di cult, time consuming, and happens while the catheters
are still in the patient. A fully automatic solution could signi cantly reduce procedure
time and potentially even improve the precision. This thesis introduces a novel
approach that segments 2D ultrasound images using the successful U-Net architecture
to determine catheter candidates. These candidates are then extracted and Fourier
Transformed into the frequency domain. De-convolution is performed directly in the
frequency domain to reconstruct a number of frequency coe cients and remove noise.
Additional features are calculated from the frequency coe cients to supplement the
determined U-Net con dence and candidate location. Altogether, the features from
each catheter candidate are classi ed by AdaBoost.Studentye