21793 research outputs found
Sort by
Discrete-Time Arbitrage-Free Nelson-Siegel Model and its Applications to Participating Life Insurance Contracts and Swaptions Pricing
Get PDFThis dissertation explores the importance of interest rate modeling in finance and actuarial science. It emphasizes the significance of yield curve modeling in pricing financial instruments, such as participating life insurance contracts and swaptions. The dissertation extends the work of previous studies and proposes a slightly different version of the discrete-time arbitrage-free Nelson-Siegel model (DTAFNS), providing a closed-form expression for risk-free spot rates and demonstrating its superior out-of-sample predictive ability. Additionally, the dissertation focuses on stochastic interest rates and mortality dynamics' impact on the pricing, reserving, and risk measurement approaches of participating life insurance contracts, with the introduction of a shadow reserve to improve accuracy. Lastly, the dissertation outlines procedures for pricing swaptions under the DTAFNS model. Overall, this dissertation contributes to the stability of the financial sector and protecting the financial well-being of individuals and institutions
A field study of thermal comfort and summertime overheating of six schools in Montreal Canada
No full textAs a result of global climate change, the world has witnessed a noticeable rise in average temperatures along with a surge in the frequency and intensity of extreme weather conditions such as heatwaves. Indoor overheating is a growing concern, particularly for vulnerable populations such as primary school students. To investigate this issue, a field monitoring network was established in six primary school buildings in Montreal, Canada. This network provided field measurements, including indoor and outdoor temperature, humidity, wind speed, and solar radiation, at five-minute intervals. This paper presents a case study that focused on three time intervals during the summer months: two school closed periods in 2020 and 2021 (unoccupied) and one school open period in 2021 (occupied). The study used an adaptive model to analyze the indoor thermal condition between unoccupied and occupied periods, non-heatwave periods, and heatwave periods.
The study concluded that natural ventilation in buildings posed a risk for indoor overheating during heatwaves, while buildings with mechanical ventilation systems had better indoor thermal conditions. The correlation analysis showed that the building's response to outdoor weather factors in naturally ventilated buildings is consistent. Multiple linear regression analysis confirmed that outdoor temperature was the most significant factor affecting indoor thermal conditions, followed by solar radiation, wind speed, and relative humidity. Furthermore, the indoor and outdoor temperature difference shows a stronger linear correlation with the indoor temperature than the outdoor temperature in all school buildings
Three Essays on Effective Social Media Marketing: Overcoming Challenges and Maximizing Opportunities in Today's Business Environment
Get PDFThe current dissertation is a collection of three essays in the context of social media challenges. As social media (SM) has become an essential tool for businesses, it has also brought new possibilities for marketers to connect with their customers. Social media marketing (SMM) has gained increasing importance in recent years, but it also poses several challenges for individuals and organizations.
Essay 1 presents a comprehensive review of the challenges of SMM by examining more than 80 publications from 2007 to 2021, categorizing them based on different metrics, and extracting a theoretical framework of common SMM challenges. This research highlights the need for decision-makers to identify and assess these challenges to better allocate resources and increase the effectiveness and efficiency of SMM.
Essay 2 focuses on the beauty industry in Iran and identifies the main SMM challenges faced by companies in this sector. Drawing on the theoretical framework developed in essay one, this study uses a Delphi survey to gather data from eight marketing and SMM managers and practitioners from various Iranian beauty companies. The findings highlight the most challenging aspects of SMM in this industry, including coordination across different company functions engaged in SMM, maintaining security and privacy of company channels and customer data, and limited IT resources of some companies in adopting SMM. The results of this study provide practical implications for beauty industry practitioners and contribute to the academic domain of SMM and business marketing practice in the beauty industry.
Essay 3 focuses on the key factors influencing the selection of social media influencers (SMIs) for small- and medium-sized enterprises (SMEs). This study uses interviews and survey methodology to identify and evaluate the key factors influencing the selection of desired SMIs for SMEs. The findings indicate that engagement, SMI traits, content relevance, compensation, and SMI workstyle are the most significant factors that positively influence SMEs' choice of SMI. This study provides a framework for future work to assess, compare, and select the most influential SMIs in different organizational contexts and helps marketers choose SMIs who resonate well with designated customers to enhance poFsitive brand association and increase product sales
MARFL: An Intensional Language for Demand-Driven Management of Machine Learning Backends
Get PDFArtificial Intelligence (AI) is a rapidly evolving field that has transformed numerous industries and one of its key applications, Pattern Recognition, has been instrumental to the success of Large Language Models like ChatGPT, Bard, etc. However, scripting these advanced systems can be complex and challenging for some users. In this research, we propose a simpler scripting language to perform complex pattern recognition tasks.
We introduce a new intensional programming language, MARFL, which is an extension of the Lucid family supported by General Intensional Programming System (GIPSY). Our solution focuses on providing syntax and semantics for MARFL, which enables scripting of Modular A* Recognition Framework (MARF)-based applications as context aware, where the notion of context represents fine-grained configuration details of a given MARF instance. We adapt the concept of context to provide an easily comprehensible language that can perform complex pattern recognition tasks on a demand-driven system such as GIPSY. Our solution is also generic enough to handle other machine learning backends such as PyTorch or TensorFlow in the future.
We also provide a complete implementation of our approach, including a new compiler component and MARFL-specific execution engines within GIPSY. Our work extends the use of intensional programming to modeling and executing scripted pattern recognition tasks, which can be used for implementing different algorithmic specifications. Additionally, we utilize the demand-driven distributed computing capabilities of GIPSY to enable an efficient and scalable execution
Engineering metabolic time-sharing in a clonal Escherichia coli population
Get PDFThe “division of labour” strategy is common among microbial communities, as dividing burdensome tasks between members of a community alleviates the strain placed on individual cells. Exploiting this phenomenon in heterogeneous microbial co-cultures for industrial synthesis of valuable compounds is limited by inefficiencies in nutrient exchange and conflicting growth requirements. Here, we demonstrate a synthetic gene circuit which enables cells of an isogenic Escherichia coli population to carry out “metabolic time-sharing” by shifting between alternate metabolic states via temporal changes in gene expression. Further, we review techniques for monitoring such dynamic processes at the single-cell level, and discuss their current applications in bacterial studies. To validate that our circuit may be used to induce cooperative behaviours in microbial populations, we adapted this circuit to engineer cells that oscillate between distinct amino acid auxotrophy phenotypes, driven by the periodic silencing of key biosynthetic genes. Culturing a clonal time-sharing population with unsynchronized oscillators permits reciprocal amino acid cross-feeding, ultimately ensuring population viability. Through comparative growth experiments, we found that the fitness of our time-sharing population was comparable to that of a heterogeneous co-culture composed of E. coli auxotrophs similarly capable of cross-feeding amino acids. Although future studies would be needed to confirm this, our preliminary results suggest that metabolic time-sharing may be a viable alternative to synthetic heterogeneous co-cultures. As it may enable an entire complex biosynthetic pathway to be engineered into a single host with reduced metabolic burden, the metabolic time-sharing strategy demonstrated here could potentially be implemented for microbial bioproduction, among other widespread applications
Load Sharing Mechanism of Micropiled-Raft Foundations in Sand
Get PDFMicropiled-Raft (MPR) foundations are economical and easy to install. In this study, the load sharing mechanism of micropiled-raft foundations in sand was examined. A series of experimental tests were performed on small-scale models in sand with different relative densities. The effects
of micropile spacing and relative density of the sand on the overall load-settlement behavior and the load-sharing mechanism were examined.
The experimental test results served to validate a series of numerical models, which were employed to produce data for a wide range of governing parameters. The effects of the micropile spacing ratio, the relative density of the sand, and the thickness of the raft were examined. While the raft stiffness only marginally affects the overall load-settlement behavior, yet, the load sharing is impacted.
Finally, the Poulos-Davis-Randolph (PDR) method was evaluated against the data produced by the numerical models. It was concluded that the PDR method was occasionally capable of determining the axial stiffness of the MPR with an acceptable range of error, however in general, it overestimated the axial stiffness of the MPRs. Thus, a modification factor was proposed which was validated by the present experimental result
Development of a Federated Learning Aggregation Algorithm
Get PDFThe Internet of Things is made possible by the recent developments in communication and 5G networks, which enable real-time sensory data transfer between billions of devices. Raw data has no value until it is processed to extract its features. The features can be extracted using a machine learning technique, a common way to use these data. Federated learning (FL) is a platform that enables a group of clients to train a model cooperatively without disclosing their personal information. Traditional federated learning has issues such as data and model poisoning attacks, free-riding attacks, and model divergence caused by clients' non-independent and identically distributed (non-IID) datasets. Because the conventional federated averaging (FedAvg) aggregation algorithm in FL lacks an evaluation technique, it is unable to detect dishonest users or correct the global model's divergence. In this study, we suggest Shapley averaging (ShapAvg), a Shapley-based aggregation technique, to aggregate the global model by analyzing the models of the clients more effectively. Each client's weight in the weighted average under this approach will be proportionate to how much it contributed to the overall model performance. The results demonstrate that while employing non-IID datasets and in the presence of data poisoning or free-riding attacks, our suggested technique overperforms the FedAvg
New Methods For Domain Adaptation And Low Data Deep Learning
Get PDFReal-world data coming from settings like hospital collections for detecting disease experience multiple sources of distributional shifts. These issues affect the performance of diagnostic methods, reducing the quality of service provided and leading to health or economic harm. Deep learning has emerged as a promising method for classification tasks, including diagnostics, and recent progress has led to methods that allow a neural network to adapt network statistics to shifts in specific settings at test time. However, problems arise in these methods adapting to general shifts and domains. In addition, they underperform when data is limited. In our first contribution, we tackle general domain shifts by investigating the key issues leading Test Time Adaptive algorithms to fail under label shift, proposing a means for mitigating these failures. In the second contribution, we tackle few-shot cross-domain adaptation by modifying the affine parameters of the batch norm during few-shot train time, generally enhancing performance. The third contribution parameterizes Scattering Networks, where we enhance a method for low data regimes by providing problem-specific adaptation
Three Essays in Microeconomics
Get PDFThis dissertation consists of three chapters that tackle topics in Microeconomics. The three essays are as follows.
Chapter 1 examines the effects of liquidity requirements on the stability of different interbank network structures. While liquidity requirements strengthen the stability of the financial system, reduce the extent of financial contagion, and prevent the failure of interbank networks, banks
with different functions within the interbank network should be bound by varied liquidity requirements. Furthermore, the study demonstrates that excessively strict liquidity requirements can impair the normal operations of financial institutions, potentially impeding economic growth.
Chapter 2 empirically analyzes the effect of the general managerial ability of CEOs on firms’ choice of successors. Using recent data collection from EXECUCOMP and Boardex, I examine the external and internal recruiting decisions of publicly traded firms in North America during the past two decades. Using an instrumental variable approach, I find that a simple probit method is likely to underestimate the effect of successors’ general abilities, while the relative importance of general
ability is lower for large firms due to asymmetric learning about internal and external candidates, as well as a trade-off between CEO ability and a significant premium to external successors or generalists.
Chapter 3, co-authored with Dr. Ming Li, theoretically studies the quality disclosure strategies of an industry in the absence of regulation and proposes the optimal disclosure strategy for an industry in a vertical differentiation duopoly model. In a price-competitive environment, we demonstrate that industries can collude on how to disclose private information about product qualities, and it is optimal for the industry to reveal only the order of the product qualities in order to maximize joint
profits. It also suggests that disclosing any quality cutoffs will not enhance the joint profit of the industry
Risk factors and prevention in the offspring of parents with an affective disorder: associations between neuroendocrine function, the caregiving environment, and child emotional and behavioural problems
Get PDFThe offspring of parents with an affective disorder (OAD) are at high risk of developing mental disorders. This thesis examines the influence of hypothalamic-pituitary-adrenal (HPA) axis functioning and the caregiving environment on the transmission and prevention of psychopathology in the OAD. First, meta-analytic procedures were used to quantitatively summarize studies comparing diurnal cortisol levels in the natural environment in the OAD to control offspring. Relative to controls, the OAD had higher mean levels of cortisol at different timepoints throughout the day (Hedges’ g = .21). These findings suggest that changes in HPA function may predate the onset of a full-blown affective disorder (AD). In the second study, data from a longitudinal study of offspring of parents with bipolar disorder (OBD) was used to study the relations between HPA axis functioning, the caregiving environment, and offspring psychopathology. As expected, the OBD who developed an AD had a higher cortisol awakening response (CAR) than OBD who did not have an AD (Cohen’s d = 0.423) and controls (Cohen’s d = 0.468). Serial mediation analyses revealed that family structure in childhood and the CAR in offspring mediated the relationship between risk status (having a parent with bipolar disorder) and offspring internalizing symptoms 12 years later (CI [.01, .66]). The last study aimed to evaluate the efficacy of the Reducing Unwanted Stress in the Home program using a quasi-experimental design with an assessment-only control group. Assessments were conducted at pre- and post-intervention, and at a three- and six-month follow-up. Multilevel modelling revealed reduced externalizing symptoms in the OBD and enhanced family organization immediately post-intervention. The gains in organization remained at the six-month follow-up, while reductions in family conflict became apparent. Mediation analyses indicated treatment induced changes in organization, but not other aspects of the family environment, were associated with reduced externalizing problems in the OBD at the six-month follow-up. Taken together, HPA abnormalities may represent a biomarker of risk among the OAD which may be shaped, at least in part, by specific, early experiences in the caregiving environment. These findings highlight the need for targeted, developmentally-informed treatments to offset adverse outcomes in the OAD