21793 research outputs found
Sort by
Learning-Based Load Identification and Forecasting using Low-Frequency Measurements
Get PDFNon-intrusive load monitoring (NILM) is a technique used for effective and cost-efficient electricity consumption management. This thesis presents two different NILM methods. One of them employs a hybrid model of convolutional neural network (CNN) and bidirectional long short-term memory (BiLSTM) for low-frequency data disaggregation, and the oher one utilizes a graph neural network (GNN) along with a long short-term memory (LSTM) network for load prediction, both combined with attention mechanism. The first study is adept at extracting temporal and spatial features from low-frequency power data, enhanced by an attention mechanism for event detection and load disaggregation. We conduct simulations using the publicly available low-frequency REDD dataset to assess our model’s performance. The proposed approach exhibits superior accuracy and computational efficiency compared to existing methods. The second study explores NILM load prediction, integrating a GNN to represent complex time correlations between appliances, forming a graph-based foundation for feature extraction. The outcome is coupled with LSTM for temporal pattern capturing and attention processes for focusing on key information. The results confirm the effectiveness of this approach in predicting load and uncovering hidden power consumption patterns. Both studies contribute significantly to the field of NILM, offering advanced methodologies for energy management in smart homes
Pedestrian Detection Systems Focusing on Occluded and Small-Scale Individuals
Get PDFPedestrian detection is essential in various applications, such as self-driving vehicles, video surveillance, and intelligent traffic management. However, the wide variations in pedestrian sizes, postures, locations, and backgrounds make the detection a complex task. In particular, the detection becomes significantly challenging due to the lack of pedestrian information when pedestrians are occluded by other objects, such as vehicles or trees, or when they appear as objects of small-scale in an input image. Such situations occur frequently in the real world. The objective of this thesis is to design CNN-based pedestrian detection models to improve the detection of occluded and small-scale pedestrians.
The first part of this work addresses the occlusion problem by proposing a specific detection model referred to as Multi-Branch Center and Scale Prediction (MB-CSP). The proposed model employs a multi-branch structure to optimize the utilization of the features extracted from the visible parts of pedestrians. This structure enables the feature data from the upper, middle, and lower parts of a pedestrian, as well as those of the full body, to be processed separately. By doing so, the data representing the true pedestrian appearances, whether partially or fully visible, can be more dominating in the final decision making. As a result, the interference from non-pedestrian data in the detection can be minimized. To optimize the fusion of the detection outcomes generated by the multiple branches, a new method referred to as Boosted Identity Aware-Non Maximum Suppression (BIA-NMS) is developed and applied in the design of the MB-CSP detection system. The BIA-NMS method eliminates redundant detections across branches and boosts the scores of the preserved detections. To implement the proposed model, a part annotation algorithm has been introduced to enable the training of the multi-branch structure. It is anticipated that the proposed model can boost the overall performance of the pedestrian detection system.
The second part of this work provides a number of approaches to improving the detection of small-scale pedestrians, besides the occluded ones. One can use two CNNs designated to detect pedestrians of large and small scales, respectively, to achieve a good detection in each of the two cases. Instead of two designated CNNs, one can use only one and incorporate a specific branch in the proposed MB-CSP model to process the features of small-scale pedestrians. The other approach proposed in this thesis is to segment the original input image into multiple partially overlapped sub-images, the likelihood of the presence of small-scale pedestrians in each sub-image is measured, and those of high scores are selected and enlarged. The detection is performed by two CNNs, of which one is designed for the original image and the other for the selected/enlarged sub-images, in order to enhance the detection of small-scale pedestrians while preserving the detection quality of the occluded pedestrians.
The detection systems presented in this thesis have been trained and evaluated using image samples from the Caltech-USA and CityPersons datasets. The tests have confirmed the effectiveness of the proposed multi-branch system in detecting occluded pedestrians. The test results have also demonstrated that the approaches to enhance the small-scale pedestrian detection produced a visible improvement in this aspect without affecting the detection of occluded pedestrians
Relating the circadian dynamics of cortical glutamate to human motor plasticity: a trimodal MRS-EEG-fMRI imaging study
Get PDFPerforming voluntary motor actions, ranging from basic movements such as walking to more complex movements like playing piano, is an integral part of our daily life. Understanding the underlying mechanism of motor learning can benefit education, sports training, and clinical rehabilitation. This study aims to investigate motor learning using diurnal variation in glutamate concentration, the main excitatory neurotransmitter in the central nervous system. To do so, glutamate concentration was measured by magnetic resonance spectroscopy (MRS) at several time points during a control visit and an MSL visit, which employed a finger-tapping sequence task. The study focused on two regions of interest: the supplementary motor area (SMA) as a motor-implicated region and the posterior cingulate cortex (PCC) as a control region.
Electroencephalogram (EEG) and functional magnetic resonance imaging (fMRI) techniques were used alongside MRS to comprehensively track the underlying neuronal mechanism that facilitates learning.
The finding of this study suggests that neuronal plasticity and the creation of memory traces rely on the modulation of glutamate concentrations, specifically in the motor-implicated area. Additionally, the results reveal a tight coupling between metabolism, cerebral blood flow, and neuronal activity, indicating the necessity of employing multi-modal imaging studies to explore learning-induced processes
White Race-Shifting: The Implications of Genetic Ancestry Testing for Critical Race Theory
Get PDFThe rise of direct-to-consumer genetic ancestry tests have implications for white racial identity formation, social capital, and race shifting, which warrants the attention of critical race scholars and cultural studies scholars. By promoting the fictional idea that DNA can reveal who we are, these direct-to-consumer devices purport to people the fantasy of transgressing boundaries of racial identity. This study uses a mixed method approach combining in-depth open-ended interviews with a qualitative relational analysis of YouTube videos to better understand how and why white people are turning to commercially available ancestry tests to race shift
Evaluation on the Effect of Mechanical Stress in Pericyclic Reactions: Mechanochemical Enabled 1,3-Dipolar Cycloaddition Between Nitrile Oxides and Alkynes
Get PDFThe synthesis of heterocycles has constituted a proliferating and growing area in chemistry. Among these, isoxazole motifs are commonly found in many drug candidates, novel materials, and versatile intermediates used to synthesize complex natural products.
Isoxazoles are typically synthesized through a 1,3-dipolar cycloaddition between nitrile oxides (NOs) and terminal alkynes. However, this type of cycloaddition often results in low regioselectivity, forming complex mixtures of 3,5-isoxazoles and 3,4-isoxazoles. To improve regioselectivity, Cu-catalysts have been shown to enhance selectivity for the 3,5-isoxazoles, while Ru(II)-catalysts favour the formation of 3,4-isoxazoles. However, these solution-based protocols suffer significant drawbacks, such as long reaction times, low atom economy, and low energy efficiency.
A more promising alternative is mechanochemistry, which offers unprecedented modes of reactivity and selectivity with a lower environmental impact. Despite the diverse application of isoxazoles, very few reports have utilized mechanochemistry to synthesize these heterocycles. Herein, we discuss the impact of mechanochemistry in combination with catalysis in the regiospecific synthesis of 3,5-isoxazoles and 3,4-isoxazoles from terminal alkynes and NOs. Furthermore, we will highlight the applicability of the developed mechanocatalytic conditions in the synthesis of trisubstituted isoxazoles from internal alkynes and NOs.
Additionally, we explored the impact of mechanochemistry in desymmetrizations by cycloaddition-type reactions, specifically, in the desymmetrization of unbiased bis- and tris-alkynes to access unprecedented 3,5-isoxazoles-alkyne adducts selectively. This approach allows for the modular synthesis of unsymmetrical bis-3,5-isoxazoles
Current Observer based Single Current Sensor Control Strategies for Permanent Magnet Synchronous Machines
No full textPermanent magnet synchronous machines (PMSMs) are widely employed in servo control and automotive applications due to their remarkable characteristics. However, the high-performance three-phase PMSM control strategies rely on accurate current measurements that require at least two phase current sensors. There is an inherent risk of current sensor malfunction which poses a direct threat to the motor drive performance and safety. Observer based single current sensor (SCS) control methods have recently emerged as practical and efficient current sensor fault tolerant control solutions, as they offer a robust and cost-effective alternative when only one current sensor is functioning in the motor drive system. In addition, the SCS control strategy can be used directly for a low cost PMSM drive design as it only requires one current sensor and achieves high performance PMSM control. This thesis delves into different current observer design methodologies aimed at achieving high-performance observer based SCS control PMSM drives with full-speed capability.
The thesis commences with a comprehensive theoretical analysis of observability and stability for observer based SCS control in PMSM drive systems, providing essential insights into feasible operation points. To enhance the system robustness and performance, an augmented linear current observer is proposed, incorporating speed-adaptive gain. Additionally, an innovative position-offset injection method is proposed to address the unobservability issue of observers during zero-speed operation. To further enhance dynamic performance, a novel nonlinear gain function is incorporated into the current observer design, facilitating rapid response and minimal overshoot simultaneously. Lastly, a hybrid discretization approach is presented, achieving a balance between simplicity and accuracy for discrete-time observer design. This approach guarantees the stability of the observer based SCS control system, even under challenging low sampling-to-fundamental frequency ratio conditions. During the thesis investigations, the proposed observer based SCS control approaches are extensively evaluated on a laboratory PMSM drive system under different speeds and load conditions.
In summary, this thesis offers a comprehensive exploration of diverse observer based SCS control strategies for PMSM drives. By scrutinizing existing observer designs, novel observer design approaches are proposed for SCS control in PMSM drives contributing to a cost effective, high performance and reliable PMSM drive solution
The Growth of Passive Investing and Its Impact on the Change Anomaly
Get PDFThis study examines the effects index changes have on securities that are added to or deleted from the S&P 500 Index from 2000 to 2019. Furthermore, it attempts to examine the relationship between the price movement of securities around index changes (as measured by abnormal returns) and the growth in indexing. I find that the increase in indexing diminishes the addition (deletion) effect, which seems to contradict the price pressure hypothesis. Vijh and Wang (2022) argued that the increased indexing or institutional ownership in the S&P 400 Index contributes to the diminishing addition (deletion) effects. Ultimately, multiple forces are clearly at work when it comes to index changes
Chaos, Desire, and the Neoliberal Self: A Socio-Theoretical Critique of Contemporary Idolization
Get PDFWith an ever-increasing engagement with the world through social media technologies, our relationship with others takes on novel directions. This thesis will approach this unusual landscape by considering both its emergence and impetus, through the dynamics of the Self. To do so, however, requires a re-evaluation of the concept of self and its critical relationship to chaos, and thereby, desire. Such will be done with the unlikely conjunction of two thinkers: George Herbert Mead and Gilles Deleuze. Despite their numerous differences, their contrasting theoretical stances will rejuvenate an image of self that is no longer a philosophic abstraction simply observing the world, but one unreservedly contiguous to natures unfolding, thus riddled with unforeseen possibilities. It will be shown that the scope of ‘our’ desire can only be captured conceptually when such a force is drawn into the core of the self, expanding the creative potentiality that lies within every individual. Such will allow us to grasp the important role celebrity figures play not only in the progression of society, but furthermore in the mediation of our very desires to a point of indolence – defining the very state of society we see today
Underdeveloped Countries in the News: The Underrepresentation and the Misinformation of the Beirut Port Explosion
Get PDFThis thesis explores the notion of underrepresentation and misinformation when looking at underdeveloped countries in the news looking specifically at the Beirut port explosion that took place on August 4, 2020. With a focus on social media, more particularly the platform of Instagram, this research is based on the fast-paced news propagation found on Instagram while noting the lack of information as well as misinformation of the Beirut port explosion. This is also related to the underrepresentation of the countries that is linked to misinformation. This thesis is merely a start as awareness is only the start for change
The Effects of Tuition Hikes on the Choices of College Majors
Get PDFThis dissertation includes an introduction, three related chapters, and a summary at the end. The primary objective of this dissertation is to explore the factors that influence college enrollment decisions and examine the impact of financial aid policies on these decisions. The thesis focuses on two distinct fields of study: STEM (Science, Technology, Engineering, and Mathematics) and ARTS (non-STEM fields defined within the thesis). STEM fields are generally associated with higher-paying career opportunities, while ARTS fields are often considered to offer relatively lower-paying career paths.
The introduction section of this dissertation begins by examining historical trends in schooling, with a focus on developed countries, notably the U.S. It discusses significant observations from the past decades, including a notable increase in college enrollment rates, the upward trend in college costs, and the presence of generous student financial aid programs. The introduction also sets the stage by posing several key research questions that will be addressed throughout the dissertation. Central to the research is the main question: How do various financial aid policies influence college enrollment decisions, particularly in the context of STEM and ARTS college majors? With these questions in mind, the introduction also outlines the unique contributions this dissertation aims to make to the existing literature.
The first chapter of this dissertation serves as an extensive literature review on educational choice. It offers an in-depth exploration of various perspectives from which scholars examine the main determinants influencing individuals’ educational decisions. Specifically, the chapter delves into how researchers have modelled schooling decisions, focusing on college enrollment and, more specifically, the college major choices, constituting this dissertation’s central theme. Furthermore, this chapter presents various educational policies implemented in the U.S., such as merit-based scholarships and need-based grants.
This chapter explores reduced-form models investigating the relationship between wages and schooling. Next, the chapter introduces the pioneers of structural models, which provide a deeper understanding of the underlying mechanisms driving educational decisions. The discussion then turns to static and dynamic empirical self-selection papers. Another essential strand of literature that this chapter covers is the research that focuses on resolving uncertainty about individuals’ tastes and abilities. Furthermore, the chapter deeply examines the subjective expectations data framework. Researchers in this field use datasets designed to elicit individuals’ expectations about future outcomes.
The first chapter also delves into historical trends in U.S. college attainments and wage premiums in the twentieth century. This chapter discusses the existing research that explains changes in the mean ability of different educational groups over time. Furthermore, the chapter explores how college attendance patterns have reversed over time, specifically examining the role played by family income and academic ability before and after World War II. Another focus aspect of this chapter is the puzzle of the higher trend of college wage premiums compared to the lower rates in college enrollment over the past decades. The literature on the determinants of college or college major choices is categorized, encompassing monetary and non-monetary factors influencing educational decisions. Additionally, the chapter explains studies investigating the impact of various financial aid policies on college enrollment rates. The chapter concludes by describing various educational policy experiments conducted in the U.S. over the past decades.
The second chapter of this dissertation focuses on establishing crucial empirical facts about working individuals in the U.S. economy. These facts will serve as essential inputs for calibrating the life cycle model presented in the following chapter. The information of interest includes income life cycle statistics such as mean, mean/median, and Gini coefficients. These statistics will be derived from the Panel Study of Income Dynamics (PSID) for the survey years 1968-2019. They are vital for mapping the life cycle model to real-world data and finding the relevant distributional moments of the benchmark model economy. These moments are means, standard deviations and cross-correlations of three initial endowments of agents after graduation from high school (learning ability, the initial stock of human capital and the initial assets). Moreover, the chapter estimates the skill price growth rates and human capital depreciation rates for three educational categories. I obtained the growth
rates of skill prices equal to 0.53%, 0.35%, and -0.13% and depreciation rates of human capital as 0.9%, 0.6%, and 0.0% for STEM, ARTS, and no-college individuals, respectively.
Chapter three of this dissertation employs a heterogeneous life cycle, a human capital model, solved using dynamic programming techniques and fitted to the sample data from the PSID. The results from this chapter reveal essential insights into the characteristics of college students and their comparative advantage in learning abilities. Specifically, the analysis shows that, on average, college students have higher learning abilities, allowing them to acquire human capital more efficiently. This comparative advantage in learning ability enables them to accumulate higher levels of human capital, which will receive higher skill prices in the labour market, leading to better career prospects. As a result, college students, especially those in STEM fields, enjoy higher wages, earnings, and consumption paths compared to individuals who do not choose to go to college.
Chapter three presents ten policy experiments based on the real-life education policies discussed. These experiments investigate the potential impact of different policy interventions on college enrollment and major choices. The policy experiments include a 30% increase in merit-based scholarships, a 30% increase in need-based grants, a 30% reduction in tuition and fees, a 35% increase in federal loan limits, three extensions to the merit-based scholarship policy, and three modifications in the need-based grants policy. The extensions entail expanding the maximum aid amount to more eligible agents, extending the financial aid coverage to more individuals, or both.
Among the various policies examined, the reduction in tuition and fees stands out as the most effective in increasing overall college enrollment and enrollment in STEM and ARTS fields. This policy significantly increases 9.9 percentage points for college enrollment, 0.3 percentage points for STEM enrollment, and 9.6 percentage points for ARTS enrollment. However, despite its effectiveness, the reduction in tuition and fees is not considered costefficient. This policy distributes financial aid broadly among all individuals without targeting specific subgroups. This result allows for exploring alternative approaches for increasing college or major-specific enrollments.
The results also underscore the superiority of policies that directly and precisely focus on specific groups of individuals instead of policies lacking a distinct target group. For instance, one highly effective and efficient experiment entails extending the eligibility for merit-based financial aid and expanding the maximum scholarship amount concurrently. This policy substantially increases college enrollment by 5.4 percentage points, demonstrating both efficacy and efficiency. Remarkably, it boosts 3.3, which is lower than a less effective policy that entails widening the threshold for receiving need-based
grants. Despite its relatively modest 1.5% increase in college enrollment, this less-effective policy exhibits higher efficiency, that is, $4.3. These findings highlight the importance of effectiveness and efficiency in addressing financial barriers low-income students face.
By identifying the most effective and efficient financial aid policies, this chapter provides crucial insights for policymakers aiming to promote college access and major choices. The findings highlight the importance of targeting specific subgroups and tailoring financial aid strategies to achieve the desired outcomes in college enrollment and major decisions.
Keywords: Human capital accumulation; Comparative advantage; College, STEM, and ARTS enrollments; Tuition hike; Financial aid policies; Dynamic programming; Simulated annealing method