21793 research outputs found
Sort by
Nonlinear dynamic modeling and model-based AI-driven control of a magnetoactive soft continuum robot in a fluidic environment
No full textIn recent years, magnetoactive soft continuum robots (MSCRs) with multimodal locomotion capabilities have emerged for various biomedical applications. Developments in nonlinear dynamic models and effective control methods for MSCRs are deemed vital not only to gain a better understanding of their coupled magneto-mechanical behavior but also to accurately steer the MSCRs inside the human body. This study presents a novel dynamic model and model-based AI-driven control method to guide an MSCR in a fluidic environment. The MSCR is fully exposed to fluid flows at different rates to simulate the biofluidic environment within the body. A novel nonlinear dynamic model considering the effect of damping and drag force attributed to fluidic flows is first developed to accurately and efficiently predict the response of the MSCR under varying magnetic and mechanical loading. Fairly accurate correlations were observed between the theoretical responses based on the developed magneto-viscoelastic model and the experimental data for various scenarios. A novel model-based control algorithm based on a fractional-order sliding surface and deep reinforcement learning algorithm (DRL-FOSMC) is subsequently developed to accurately steer the magnetoactive soft robot on predefined trajectories considering varying fluid flow rates. A fractional-order sliding surface and a compensator, trained using the deep deterministic policy gradient algorithm, are designed to mitigate the amount of chattering and enhance the tracking performance of the closed-loop system. The stability proof of the developed control algorithm is also presented. A hardware-in-the-loop experimental framework has been designed to assess the effectiveness of the proposed control algorithm through various case studies. The performance of the proposed DRL-FOSMC algorithm is rigorously assessed and found to be superior when compared with other control methods
The Effect of Gender and Funding on Research Performance
Get PDFIn spite of various improvements and increasing involvement of female researchers in scientific
activities in recent years, the gender gap still persists and women remain greatly
underrepresented in technology, engineering, and computer science fields. This thesis attempts
to shed some light on the effect of gender and funding on research output of Canadian
researchers in natural sciences and engineering. In this research, using NSERC and Scopus
data from 1982 to 2018, we apply descriptive statistical analysis and regression analysis to
study the influence of funding and gender on the quantity of published journal papers and their
scientific impact. The study concludes that funding has a positive impact on both the number
of papers published and the number of citations received by their respective author. However,
we also observe that as career age of authors increases, researchers become less productive,
they publish less papers and their citation counts slightly diminish with time as well, even
though their funding amounts typically increase. In terms of gender, even though we find that
female researchers are indeed greatly underrepresented and receive lower amounts of funding
than their male counterparts, they produce on average similar number of articles with similar
scientific impact. This means that female researchers can generate comparable research output
with lower research costs compared to male researchers and are thus more efficient in their
research production. These findings suggest that governmental funding agencies should
introduce more effective gender-related funding strategies and greater support for early-career
researchers
A First Principles Study on The Effect of Hydrogen on The Electrical Properties of BCC Fe
Get PDFIron (Fe) and iron-carbon (Fe-C) materials, used in various critical industrial applications, are susceptible to Hydrogen Embrittlement (HE) – a phenomenon leading to increased crack growth rate, reduced ductility, and potentially leading to catastrophic material failure. The intricate mechanisms underpinning HE demand comprehensive scrutiny. A pivotal aspect often overlooked is how hydrogen assimilation in these materials induces alterations in specific physical attributes, notably electrical resistivity. Addressing this gap, the study employs Density Functional Theory (DFT) and the MD-Landauer method to delve into the influence of hydrogen on the electrical characteristics of Fe and Fe-C systems. The study shows that the presence of interstitial hydrogen increases the resistivity of Fe and Fe-C systems. Hydrogen is seen to reduce the likelihood of electron transmission in the system by disrupting the electron density distribution. This introduces a noteworthy increase in resistivity with hydrogen incorporation. Moreover, the presence of defects, like vacancies and grain boundaries, is shown to increase the resistivity of these systems. Adding hydrogen along with these defects further increases the resistivity. Studying the effects of hydrogen on the physical properties of metals may pave the way for non-destructive detection of hydrogen presence in metals. Such understanding is crucial for enhancing quality control, ensuring material longevity, and preventing potential failures during manufacturing processes or in-service
Blood Pressure Estimation through Photoplethysmography using Deep Learning in Clinical Setting: Critical Survey and Solutions
Get PDFCurrent solutions for blood pressure monitoring can be classified as invasive or non-invasive,
both with drawbacks. Invasive blood pressure monitoring can lead to complications. Non-invasive
blood pressure monitoring is intermittent which leads to missed episodes of hypertension and hy-
potension, also leading to complications. The state of the art for blood pressure monitoring through
machine learning methods usually requires personalization, which is prohibitive in a clinical appli-
cation. These proposed methods are generally not evaluated for clinical application. Datasets are
usually split randomly, while a patient-wise split is required.
We first start by performing a survey of the literature to find candidate models for evaluation.
These models are reproduced for evaluation alongside our proposed models. Popular input modal-
ities from the literature are also reproduced with our proposed input modality. All combinations of
models and input modalities are then evaluated against a patient-wise and random split. We perform
a learning curve analysis to estimate how much data would be required to pass the AAMI standard.
The performance results establish that no model can provide calibration-free, non-invasive blood
pressure monitoring using a single PPG site. The performance metrics show that our models and
input modalities outperform the state of the art for random and patient-wise splits. Comparison
against the models demonstrates that model complexity is insufficient to achieve better performance
and that better preprocessing is a more efficient way to improve performance. The learning-curve
analysis estimates that additional data could help achieve a model that passes the AAMI standard
Three Essays on R&D Competition with Spillovers: Theory and Experiment
Get PDFThis thesis consists of three chapters. The first chapter reports a laboratory experiment on dynamic patent races in an indefinite horizon with complete information. In the experiment, we examine how the players react to a leader/follower or symmetric/asymmetric position as well as the distance between the initial knowledge stock and the target. Our results show that the individual average effort is highest for the players who are in a tie position, second highest for the leaders and lowest for the followers and the spillovers in the previous round significantly increase the players’ investment in the current round. By comparing the first and second half of the session, we observe an overall learning effect on the pure-strategy equilibrium play, but efficiency loss remains throughout the session.
The second chapter investigates the effect of R&D subsidies on the innovating firms’ quality investment choices and profits as well as social welfare in a duopoly market with product substitutability, demand spillovers and consumers’ quality sensitivity. Taking the non-cooperative and cooperative scenarios into account, the optimal R&D subsidy levels are solved in a way to maximize the social welfare. Compared with no-subsidy, the firms are better off under the R&D subsidy policy. Furthermore, it is always socially beneficial to subsidize the non-cooperative regime or the cooperative agreement.
The third chapter considers a two-stage strategic R&D model in a duopoly market. In the first stage, two firms decide simultaneously whether to compete or to cooperate by choosing the level of R&D investment that might decrease the investing firm’s production cost and the rival’s cost through the absorptive capacity. In the second stage, after observing the R&D outcome, the two firms play the classical Cournot in order to maximize their own profits. Under the stochastic R&D technology with low or high symmetric absorptive capacity, I find that the difference between the optimal R&D expenditures under defection and those under cooperation becomes larger as the probability of success increases. Regardless of whether the absorptive capacities of the two firms are same or different, except at the critical threshold, the R&D outcomes always align with the prisoner's dilemma situation
Performing Blackness/Archiving Whiteness: The Y Minstrels, 1927-1951
Get PDFThe province of Québec is often not included or under-studied in research regarding blackface minstrelsy, but it would be fallacious to assume that its underrepresentation in publicly accessible archives equates its non-existence. Blackface minstrelsy is a racist theatrical form in which white actors and musicians, the minstrels, would apply burnt cork to their faces and perform derogatory stereotypes of Black people in front of white audiences. As Cheryl Thompson argues, white Canadians reproduced American minstrelsy, not to soothe class fears, but in response to their fears about Black immigration and their supposed inability to assimilate into Canadian culture. Looking specifically at Montréal, blackness was used as a satirical tool to debase francophone bodies, which were deemed “other” to the Anglo-Saxon English-speaking majority at the time. Henri Julien’s Songs of the By-Town Coons exemplifies how blackface minstrelsy imagery found its way into mainstream media and the width of its popularity.
Located in the Jewish Public Library archives, the Young Men/Women Hebrew Association (YM-YWHA) and the Irving Silverman fonds contain a plethora of photographs and documents related to their very own minstrels, which were part of their musical programme. Examining the YMHA minstrels as a case study allows for the first in-depth analysis of this troupe and for situating it within the broader context of Canadian blackface minstrelsy. This project then observes the role of the performances and their recording in the Jewish community’s own archives as a necessity for self-representation while simultaneously highlighting the pride they took in their performances
Three Essays on Cryptocurrency Analytics and Forecasting by Using Deep Learning Models
Get PDFAbstract
Three Essays on Cryptocurrency Analytics and Forecasting by Using Deep Learning Models
Bahareh Amirshahi, Ph.D.
Concordia University, 2024
Since the emergence of Bitcoin in 2008 as the first cryptocurrency, digital assets have become favored investment options worldwide. Understanding and predicting the behavior of
cryptocurrency markets are essential for effective risk management and investment decisions.
However, the rapid fluctuations in these markets make accurate predictions a challenging task. In
this thesis, key aspects of cryptocurrency market analytics are explored from three different angles.
The recurring theme in each study involves proposing hybrid prediction models by combining a
feature extractor component with deep learning models to enhance prediction performance.
The first study focuses on predicting cryptocurrency volatility, an underexplored area
despite extensive studies in financial markets. By combining traditional econometrics methods
with deep learning models, we forecast daily volatility with improved accuracies. The findings
revealed that deep learning models not only enhance the accuracy of traditional models but also
exhibit superior forecasting when combined with such models in a hybrid approach.
In the second study, we address the challenge of predicting cryptocurrency price values.
Recognizing the impracticality of a universal model due to unique cryptocurrency characteristics,
we propose a flexible architecture tailored for each cryptocurrency. Additionally, we explore the
impact of sentiment data from Twitter posts on prediction accuracy, employing state-of-the-art pretrained
language models in an ensemble manner for more robust sentiment analysis. We show that
sentiment data improves the prediction results for more than 70% of the cryptocurrencies studied.
This indicates that social media posts, in particular tweets play a significant role in the
cryptocurrency markets behavior.
iv
The third study focuses on predicting the direction of cryptocurrency prices. We propose
an innovative approach by combining data denoising techniques with machine learning methods,
that generates high-quality data for prediction models and achieves significantly higher accuracies.
Notably, we assess the proposed approach across distinct periods: before, during, and after the
COVID-19 pandemic, filling a critical gap in research regarding predictive models during crisis
periods. We show that the predictive performance of the proposed method is not affected by high
values of volatility in challenging periods like the COVID-19 pandemic.
This research helps in developing highly accurate prediction models that can aid investors
seize profitable opportunities and avoid potential losses. By analyzing over 25 cryptocurrencies,
collectively representing 75% of the total market capitalization, this study offers a comprehensive
perspective beyond the conventional Bitcoin-centric approach.
Keywords: Cryptocurrency, Forecasting, Machine Learning, Deep Learning, COVID-1
Membrane Protein Classification with Protein Language Models
Get PDFThis thesis investigates the application of Protein Language Models (PLMs) to enhance the classification of membrane proteins, which are crucial for cellular functions and pharmacological targeting but challenging to characterize due to their context within a membrane. We employ PLMs derived from Large Language Models of natural language processing, including ProtBERT, ProtT5, ESM1b, ESM2, and Ankh. These PLMs are pretrained using self-supervised learning on extensive datasets such as UniRef50 (40 million proteins) and BFD (2 billion proteins).
Our research comprises four interconnected projects focused on discriminating membrane proteins, transport proteins, and ion channels from proteins not in those classes. We use established state-of-the-art (SOTA) tools with standard datasets for training and testing as a baseline for evaluating our work.
The first project demonstrates that fine-tuning is beneficial in classifying membrane proteins, with a fine-tuned combination of ProtBERT-BFD and logistic regression (LR) outperforming SOTA. The second project shows that Convolutional Neural Networks (CNNs) are superior to traditional classifiers when used with PLMs for membrane protein, transport protein and ion channel classification, again surpassing SOTA performance.
In the third project, we evaluate six PLMs and six downstream classifiers across three tasks, considering fine-tuned and frozen representations, dataset balance, and floating-point precision. ESM-1b emerges as the top performer across most tasks and metrics. We confirm that fine-tuning outperforms frozen representations, imbalanced datasets work best, and there is no statistically significant difference between half- and full-precision computations.
The fourth project incorporates secondary structure information into Ankh. Evaluation across multiple tasks shows little statistically significant difference between Ankh and the modified PLM with secondary structure information.
The tools developed in this research now represent the state-of-the-art in membrane protein classification. Our methodological findings provide insights into PLM applications for protein classification in general, with particular relevance to membrane proteins highly relevant to drug discovery
The in-flight oxidation of MCrAlY particles in the HVAF process: A numerical study
No full textMCrAlY powder where M stands for Nickel (Ni) or Cobalt (Co) is extensively used in the aerospace industry as a bond coat for its corrosion and oxidation resistant characteristics. MCrAlY can sustain prolonged exposure to high temperatures such as in gas turbines. The aluminum present predominantly in the β-phase of the alloy forms a protective oxide called aluminum oxide or alumina. This latter acts as a barrier for further oxidation provided that there is no chemical breakdown resulting in the formation of non-protective oxides.
Numerical analyses have been made in place to predict the degree of oxidation of the feedstock during their flight in thermal spray processes. The most prominent is the Mott and Cabrera theory developed in 1944. The theory stipulates that the film oxide can be associated with three growth processes/thicknesses namely, very thin films, thin films, and thick films. The in-flight oxidation of particles is proven to be in the order of milliseconds, hence only very thin films are formed. In this range of thickness, the oxidation is driven by the presence of a strong electric field.
High velocity air-fuel (HVAF) has revolutionized the very exclusive group of solid-state deposition. The process comes in with the promise of reducing the particle temperature while increasing their velocity with the stated goal of diminishing the oxidation and enhancing the adhesion particle-substrate. The combination of HVAF and MCrAlY delivers qualitative coatings with low oxide content and a longer lifespan. The numerical modeling of the in-flight oxidation of the particles is validated by examining the oxide layer thickness using a Transmission Electron Microscope (TEM)
Corps mobiles et connectés : design pédagogique pour la création multimodale dans les réseaux sociaux
Get PDFLes adolescent·e·s d’aujourd’hui évoluent au contact des technologies mobiles et de réseau qui leur permettent de socialiser et de partager du contenu à tout instant, en tout lieu. Bien que cette hyperconnexion soit de plus en plus étudiée par la recherche, il ne semble pas y avoir de consensus sur ses bénéfices et ses revers. Ainsi, cette thèse se penche sur une réponse critique à cette problématique dans le champ de l’éducation artistique. Par le détournement technologique, cette recherche intègre ces technologies au cœur d’une intervention pédagogique en classe d’art au secondaire afin de mobiliser la corporéité des adolescent·e·s pour la création artistique. Son cadre conceptuel se déploie autour des notions de coprésence virtuelle continue, de spatialité et de mobilité appliquée à l’éducation, de détournement, de pédagogie de la corporéité et de multimodalité.
Au moyen de la recherche design en éducation, cette recherche a été menée avec quatre groupes et deux enseignantes en arts plastiques et multimédia dans deux écoles secondaires de la grande région de Montréal. Centrale au design éducatif, la corporéité de l’élève y est entrevue comme une tactique. Des missions de création performatives ont invité les élèves à produire et à publier une multitude de créations numériques multimodales au sein d’un réseau social éducatif en circuit fermé. Ces missions leur ont permis de créer en prêtant attention à leur corps, à leur sensorialité, à leur environnement ainsi qu’à leurs pairs. Les données de recherche résident dans les créations numériques multimodales des élèves ainsi que dans des entretiens avec les élèves et les enseignantes.
Si la pandémie de Covid-19 a ramené à l’avant-plan la saturation des pratiques en ligne des jeunes, les principaux résultats de cette recherche apportent un éclairage nouveau sur des rapports sensibles qu’ils peuvent y vivre. Ces résultats démontrent que l’usage des technologies mobiles et de réseau en classe d’art, organisé autour de la corporéité, donnent la possibilité aux élèves d’être présents à leur environnement et d’observer, de ressentir, de réfléchir et de créer d’une manière renouvelée. La mobilisation du corps opère un renversement de la démarche de création artistique habituellement vécue en classe d’art en mettant de l’avant la prise de risques et la multiplication des créations numériques multimodales spontanées. La coprésence virtuelle continue, qui caractérise l’hyperconnexion des adolescent·e·s, participe à une expérience artistique qui les met en contact avec l’univers personnel de leurs pairs. Dès lors, chaque élève délimite les frontières de son intimité et fait des choix sur ce qu’il ou elle souhaite montrer. Enfin, ce type d’intervention contribue à rapprocher l’enseignement des arts plastiques et multimédia au secondaire des pratiques artistiques actuelles tournées autant vers les processus que les productions.
Today’s teenagers are growing up with mobile and network technologies that enable them to socialize and share content anytime and anywhere. While this kind of hyperconnection is increasingly studied by research, there is no consensus on its benefits or drawbacks. Thus, this dissertation focuses on providing a critical response to this issue in art education. Through a technological détournement, the research integrates these technologies into an educational intervention within secondary school art classes to activate teenagers’ sense of corporeality in art making. It is a conceptual framework that revolves around continuous virtual co-presence, spatiality and mobility applied to education, détournement, embodied pedagogy, and multimodality.
Using design-based research, the study was conducted with four groups and two visual arts teachers in two secondary schools in the Greater Montréal area. The students' corporeality is central to educational design. Performative creative missions invited students to produce and post many multimodal digital compositions within a closed educational, social network. The missions awakened them to their bodies, sensoriality, environment, and peers. The students’ multimodal digital compositions, as well as interviews with students and teachers, make up the data of this research.
While the COVID-19 pandemic has brought to the forefront the saturation of teens’ online practices, the main findings of this research shed light on new and significant experiences they may have online. This study found that using mobile and network technologies in the art classroom that centers embodied pedagogy helps students awaken to their environment and allows them to observe, sense, reflect and make art in a renewed way. The mobilization of the body challenges the typical art-making process experienced in a class by promoting risk-taking and increasing spontaneous multimodal digital compositions. The continuous virtual co-presence that characterizes teenagers’ hyper-connectedness contributes to an art experience that connects them to the personal universe of their peers. Each student defines the boundaries of their intimacy and chooses what they wish to share. This type of intervention brings visual art education in secondary schools closer to contemporary art practices that are as much process-oriented as production-oriented