IRis Bishop's University
Not a member yet
114 research outputs found
Sort by
Prediction of Resource Demand for Virtual Machines in Clouds
Full text linkThe emergence of cloud resources represents a gradual and powerful evolution in computing, stemming from a convergence of technological advancements. In fact, the rise of the cloud is a direct response to the limitations and inefficiencies of traditional self-hosted IT infrastructures. These resources include a wide range of on-demand services such as computing power, storage, databases, networking, and software delivered over the Internet. Cloud computing infrastructures are highly dependent on accurate forecasting of virtual machine resource demands to ensure optimal performance, cost-efficiency, and service quality. The prediction of cloud resource utilization is a critical determinant of an organization’s operational costs, system performance, and service reliability. Forecasting future resource usage facilitates organizations in mitigating the risks associated with both excessive resource allocation (over-provisioning) and insufficient capacity (under-provisioning). To predict cloud resource demand, this paper introduces a hybrid deep learning architecture for forecasting cloud resource consumption, integrating Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM), Gated Recurrent Units (GRUs), and attention mechanisms. A comprehensive evaluation was conducted on the Bitbrain dataset, comparing the proposed model against established architectures, including LSTM, GRU, and others. The results indicate the notable performance improvements of the proposed model, particularly in its ability to accurately predict peak resource consumption, which is critical for maintaining cost-effective and reliable cloud infrastructures. For the specific task of CPU consumption forecasting, the model achieved an 2 score of 0.93, representing a 27 percent improvement over a standard LSTM model. In addition, an ablation study and a feature importance analysis were performed. The study revealed that convolutional layers with different filter sizeswere the most significant contributors to the high accuracy of the model. The feature analysis identified memory-related metrics as the most influential predictors of CPU consumption, while network-related metrics had the least impact.© Mitra Kezli, 202
Single-Trial ERSP Classification of EEG Data Across Multiple Axes Using Deep and Machine Learning: Spectrograms vs. Frequency- and Subject-Level Analysis
Full text linkThis study presents a novel approach for classifying visual stimuli from single trial Event-Related Spectral Perturbation (ERSP) responses derived from electroencephalographic (EEG) recordings. To improve the reliability and interpretability of visual Brain-Computer Interfaces (BCIs), we propose a classification approach that systematically evaluates neural responses across five key dimensions: subject variability, frequency bands, trial-level differences, stimulus categories, and post-stimulus time windows. To address the multi-dimensional nature of ERSP data, both deep learning and classical machine learning approaches were applied to EEG recordings from thirtyone neurologically and psychiatrically healthy participants. Deep learning models— including convolutional neural networks (CNNs) and recurrent architectures such as bidirectional long short-term memory (LSTM) and gated recurrent units (GRUs)—were trained on spectrogram representations of ERSPs. A CNN architecture incorporating dropout and batch normalization achieved validation accuracies of up to 85.57% in binary classification tasks involving high-contrast stimuli and participants with strong gamma-band responses. However, performance declined in multiclass tasks involving all stimuli and participants, with validation accuracy plateauing around 30%, primarily due to substantial inter-subject variability and limited training data.
In parallel, classical machine learning approaches were applied in a subject and frequency-specific manner. Random Forest classifiers delivered strong and interpretable results, particularly in the gamma frequency range (20–40 Hz), early post-stimulus intervals, and among participants with pronounced gamma activity. Ensemble-based models consistently outperformed support vector machines, logistic regression, and gradient boosting, underscoring the effectiveness of individualized modeling in data-constrained settings.
This study is among the first to systematically investigate ERSP classification across multiple predictive dimensions using both deep spectrogram-based models and subject- and frequency-aware classical approaches. The findings emphasize the complementary strengths of these methods: deep learning performs well with structured, high-resolution inputs when sufficient data are available, while classical approaches offer robustness and interpretability in subject-specific contexts. Future research should explore hybrid strategies, subject-informed training methods, and the application of self-supervised or transfer learning approaches fine-tuned for task-relevant scenarios to compare and investigate the possibility of improving generalizability across participants and experimental conditions.© Arian Yavari, 202
“This Work Is Amazing… Keep Doing It” A Community-Engaged, Trauma-Informed, Mixed-Methods Inquiry Assessing Trauma-Awareness Training Tailored With Work-Integrated Learning Practitioners
Full text link© Mélanie Letendre Jauniaux, 202
Native English-Speaking Teachers’ Perceptions of Language and Cultural Issues in English Classrooms in China
Full text linkIn recent years, global demand for English as a second language (ESL) education has surged, especially in non-English speaking countries like China. This demand has often led to a preference for Native English-speaking teachers (NESTs) based on biases linked to nationality, accent, and appearance. However, to meet this demand, private educational institutions have relaxed accreditation requirements, allowing recruitment of underprepared teachers through commercial agencies. Some agencies exploit visa loopholes, hiring individuals on student, tourist, or spouse visas, who might not have the proper training needed to teach.
This influx of these underprepared teachers presents challenges for both the local ESL education system and the teachers themselves, who often struggle to adapt to the new teaching cultures and exam-centered practices. This study focuses on five NESTs who worked in China, each with varying levels of training in ESL/EFL (English as a Foreign Language). It explores their attitudes towards bilingual teaching, especially regarding dynamic bilingualism in the classroom. The findings show that most participants were drawn to teaching abroad by the professional opportunities and their associated financial gains as well as the chance to explore a new culture. The high demand for NESTs allowed them to secure positions with minimal preparation, resulting in difficulties adjusting to local educational expectations. Despite these initial struggles, all participants experienced professional growth and came to value incorporating students’ native language and culture into ESL/EFL lessons.
This research highlights the importance of training in dynamic bilingualism and cultural awareness for NESTs. While NESTs can bring valuable contributions to ESL/EFL education, proper training and ongoing professional development are essential. Such teacher preparation not only equips teachers with the skills needed to address students’ learning needs but also fosters culturally responsive teaching practices. This study emphasizes the need for comprehensive teacher development to support NESTs in their roles abroad.@ Chang He, 202
Pupillary Response Function to Spontaneous Fluctuations in EEG Alpha Power
Full text linkEEG alpha rhythm (8-13 Hz.) is thought to provide an index of brain ‘idling’ or relaxation, with increases in alpha power associated with increased relaxation. Conversely, decreases in alpha power or the ‘alpha desynchronization’ are associated with increased attention or engagement of the brain. Similarly, if ambient light is kept constant, pupil diameter changes can serve as an index of attention, with dilated pupils indicating increased attention or excitement, and constricted pupils indicating decreased attention. Despite the similarity in response of pupil diameter and EEG alpha power to changes in attention, little is known about how pupil diameter and EEG alpha power co-vary under spontaneous recordings. In particular, it is not known whether changes in pupil diameter precede changes in alpha (or vice versa) and what is the time lag of the correlation between the two signals. Here, we present for the first time results from a study examining in 73 healthy subjects the relationship between spontaneously recorded pupil diameter and alpha power. Our study investigated the relationship between alpha power and pupil diameter in younger and older adults during passive, go/nogo, and simple reaction time tasks. Results indicated that both age groups had high pupil diameters at alpha power peaks, followed by a decrease. Younger participants exhibited higher and quicker dilation peaks compared to older adults, with a pupil diameter decrease lagging 1-2 seconds after the alpha peak. Cross-correlation analysis showed that younger adults had more positive correlations at specific time lags, indicating a quicker and stronger relationship between alpha power and pupil diameter, with a positive correlation around 250-300 milliseconds, whereas older adults exhibited more delayed and weaker correlations.© Masoud Majidi, 202
How do Indigenous Youth Want to Make their Mark?: Exploring the Resurgence and Generative Possibilities of Indigenous Youth
Full text linkToday, an increasing number of Indigenous youth are embracing a movement known as resurgence, seeking to reclaim and regenerate their Indigenous cultures (Corntassel & Hardbadger, 2019). Concurrently, they are assuming leadership roles in movements like Idle No More, Land Back, and Indigenous and Black Lives Matter, demonstrating generativity, a concept traditionally associated with midlife but now recognized in younger populations (Blanchet-Cohen et al., 2023; Erikson, 1969; Lawford & Shulman, 2024). In response to this phenomenon, the Students Commission of Canada initiated the Make Your Mark conference, focusing on youth generativity and reconciliation. This study presents findings from two years of the Make Your Mark conference. The conference's first year explored youth experiences through surveys, questionnaires, and focus groups, revealing nuances in Indigenous youth's expression of generativity. Building on these insights, the second year delved into Indigenous youth's resurgence within reconciliation spaces. Through activities like sharing circles and photovoice, it became evident that Indigenous youth engage in cultural practices to redefine their identities, revitalize their cultures, and ensure its survival. This research highlights the pivotal role of Indigenous youth in cultural preservation and social change within spaces of reconciliation, emphasizing the importance of understanding and supporting their resurgence and generative capacities.© Megan Legare, 202
Analysis of Usability of Wearable Device Metrics as Predictors for Brain Phenotypes from Correlations
Full text linkIn the modern healthcare system, tracking a person’s health can sometimes be
expensive and time-consuming. Photoplethysmography (PPG) is a fairly easy noninvasive
method to get some insights about a person’s cardiovascular health. Some
research has already been done with PPG in order to predict some properties
of someone’s brain from the signal. However, even if it could be possible to avoid
complete brain scans to know more about our brain using PPG, it would still require
a trip to the hospital. Furthermore, even if small household machines are capable
of acquiring PPG data, they can still be quite expensive and non-practical for the
users. If possible, using normal, consumer-grade, wearable electronic devices to
detect anomalies or brain properties with the help of PPG could greatly benefit the
healthcare system and its users by being less expensive, less time-consuming, and
able to track the users’ health throughout their day.
For this study, 52 subjects wore a consumer-grade smartwatch for about a month,
recording the PPG data and accelerometer data. They also got a magnetic resonance
imaging scan (MRI) of their brain. The smartwatch signals were then processed and
the MRI images were segmented to compute the surface area, gray matter volume,
and cortical thickness of some regions of the brain. The data from both sources
were then compared to find correlations. Some correlations have been found in the
data. The most relevant were between the heart rate and precentral gyrus, between
the heart rate variability and parahippocampal gyrus, between the heart rate and
the inferior parietal lobule, and between the heart rate variability and the lateral
orbital cortex. Although some expected results were not present, or not as good
as they could have been, some results were promising, which is encouraging for
future research in this field. It would be interesting to perform this study using a
smartwatch with a more precise sensor and to track the changes in the subjects over
a longer period. It could also be interesting to test using machine learning.© Olivier Demers, 202
Promoting the mental health and well-being of vulnerable youth through art
Full text linkBackground: Children in rural communities of Quebec represent some of the most vulnerable populations in Canada, which has implications for their mental health and well-being, particularly regarding their access to mental health services. Indeed, conventional therapies, which are typically used to address mental health concerns, are empirically validated but less accessible. Art-based interventions, by contrast, are much more accessible, but not as empirically validated. Aim: The purpose of this project was to offer an art-based intervention for children benefiting from an after-school program in rural Quebec. Methods: 27 children (ages 10-12) in Quebec, Canada, took part in a 7-week art-based intervention. A descriptive qualitative design was implemented to examine the implications this intervention had on students, particularly their mental health and well-being. Results: The majority of students reported enjoying the intervention, and many students felt as though it had a beneficial impact on their lives. Students reported feeling positively during art-making and expressed interest in taking part in follow-up research. Discussion: Positive experiences with the present intervention support existing literature regarding the effectiveness of art-based methods for youth populations. Further research is warranted to investigate how art-based interventions, although sometimes challenging, can have favourable implications for youth through pushing personal limits.© Kyra Simons, 202
"It made me realize I was more part of Fitch Bay than I expected": Artfully Exploring Community Engagement and Belonging in Rural Quebec
Full text linkEngagement and a sense of belonging are vital for sustaining a community, particularly in a time of societal division and the Covid-19 pandemic isolation. This thesis explores how community connection in a rural village can be inspired and promoted through collaborative artmaking in a community-based art education (CBAE) mural project. Four participants from Fitch Bay, Quebec, collaboratively designed and painted a mural over the course of seven weeks. The project employed an arts-based educational research (ABER) methodology and CBAE conceptual framework. This study suggests that creating a public space mural in a rural community as an interpretive expression of the community can enable its members to connect through participation and interaction with the art. The art itself also remains in the community as a testimony to this connection.@ Gretchen Hatfield, 202
Freshwater Effects of Pesticide Mixtures in Mesocosms of Great Pond Snails (Lymnaea stagnalis)
Full text linkPesticide exposure in the environment accumulates through soil, water, and living organisms. Environmental factors such as weather and water conditions can influence how pesticides can impact an ecosystem. Pesticide mixtures from agricultural runoffs could have complex interactive effects between compounds on organisms in natural waterways. Toxicity of pesticides can be increased through additive or synergistic effects in a mixture.
This study assessed the chronic toxicity of a pesticide mixture on the great pond snails (Lymnaea stagnalis) over a 28-day period using semi-natural outdoor mesocosms. We used the mesocosms to house a natural biofilm (periphyton) that included other small invertebrates. This design incorporates trophic exposure of pesticides, and potential community effects of the pesticides on the food web. L. stagnalis is a model organism in ecotoxicology. It has well-studied growth, reproductive, behavioral, and biochemical responses to anthropogenic stress that we utilized for this study. We chose to expose the snails and biofilm to a pesticide mixture commonly found in the water surrounding corn/soybean farming of southern Quebec, Canada, based on a report by the ministère de l’Environnement et de la Lutte contre les changements climatiques de la Faune et des Parcs (MELCCFP). Our final selection included the herbicides atrazine, metolachlor, and glyphosate, as well as the insecticides clothianidin, thiamethoxam, imidacloprid, and chlorantraniliprole. The pesticide mixture varied at three environmentally relevant concentrations (atrazine: 0.5-5 ng/mL; metolachlor: 1-10 ng/mL; glyphosate: 1.5-15 ng/mL; clothianidin, imidacloprid, thiamethoxam, chlorantraniliprole: 0.05-0.5 ng/mL). We showcased physiological effects with respect to nominal reductions in pesticide mixture concentrations. We also confirmed environmental exposure concentrations in the experimental
media and tissues to provide evidence for policy makers and agronomists to make decisions about the future of agricultural practices.
Results showed that snail survival rates were unaffected by pesticides. Smaller snails had lower survival at the highest mixture concentration compared to larger snails, leading to a larger increase in the remaining 1x mixture treatment population’s average body mass. This increase in body mass could potentially be due to community effects, as shown in other studies. The effect sizes of changes in population body mass did decrease with reduced pesticide mixture concentrations. Clutch sizes increased in 1x treatments at greater rates compared to controls. However, 0.5x treatments exhibited reduced clutch sizes compared to controls. Stress-induced reproduction of L. stagnalis has been observed in other studies. However, it was unclear if the varied reproductive effects are a gradient of toxicity with respect to the pesticide exposure concentration.
Snail tissues had higher amounts of s-metolachlor, chlorantraniliprole, glyphosate, and imidacloprid than the water, indicating bioaccumulation. Biofilm tissue bioconcentrated chlorantraniliprole, s-metolachlor, and glyphosate at larger than the water. Concentrations of herbicides (atrazine, s-metolachlor, and glyphosate) in snails was significantly related to the concentrations in biofilm, suggesting possible trophic transfer. A two-fold reduction in pesticide exposure did not reduce the accumulation factors of s-metolachlor and imidacloprid in pond snails. Likewise, the concentration factors of s-metolachlor and chlorantraniliprole in biofilm were not diminished. In conclusion, pesticide mixture effects are complex, with imposed sub-lethal and community effects that can appear to benefit an organism. The costs of body mass or reproduction changes need to be studied further. Additionally, the threat of pesticide bioaccumulation in snails and its potential to move throughout the food web remains an unanswered concern.© Jared Sparr, 202