21793 research outputs found
Sort by
Quadratic Hedging in a Non-causal AR(1) Model
Get PDFThis thesis explores hedging strategies for European-type derivatives under the non-causal AR(1) Cauchy model. Recently, such non-causal models have raised much attention in the finance literature due to their ability to replicate bubbles often observed in the cryptocurrency market, as well as their tractability for pricing standard European options. However, these discrete-time models are incomplete, meaning that it is impossible to perfectly replicate a derivative's payoff in such a market. This thesis explores the use of quadratic hedging approaches to manage the risk of a derivative trader
Leveraging Machine Learning to Investigate the Impact of NSERC Funding Programs on Research Outcomes
Get PDFThis research examines the impact of various funding programs by NSERC on research outcomes. We utilize statistical models and machine learning algorithms trained on the integrated database of researchers’ publications and funding to determine the efficacy of NSERC funding programs. We aim to evaluate the effectiveness of different strategies defined by NSERC through funding programs and analyze the impact of various factors. We seek to enhance our understanding, with the aspiration that it will inform the design of more effective programs in the future.
We compare the results of linear regression, random forest, and neural networks. Then, we perform SHAP analysis to identify the most important features within funding programs. We aim to gain insights into the impact of receiving funding through different programs on research outcomes.
We observed that random forest model outperformed the other models for all dependent variables, i.e., future productivity, quality of the publication, and future co-authorships. Subsequently, we examined the significance of independent variables in predicting dependent variables across the funding programs.
For Canada Research Chairs recipients, the impact of their prior work holds greater importance in shaping research outcomes, underscoring a distinctive emphasis on research excellence within this program. In contrast, the impact of career age is lower compared to other programs. Interestingly, within the Discovery Grants program, career age becomes notably influential in predicting future productivity in favor of young researchers. Furthermore, we found an intriguing exception for researchers with a history of large group collaborations within Discovery Grants, where some experience a negative impact on future collaborations. The award amount plays a more important role in shaping the research outcomes of recipients engaged in strategic projects.
Our findings emphasize the importance of allocating funding programs to researchers whose qualifications are aligned with the programs’ objectives
Enhancing understanding of experimental designs: treatment levels and choice of analytics to improve statistical performance for ecological experiments
Get PDFExperimental design is a fundamental component of research in ecology and other disciplines. It is critical to understand the consequences of statistical inference, including power and effect size, when making decisions about designing experiments. However, issues such as file drawer effect, funding logistics, and reproducibility are a major concern that often are not considered when starting one’s scientific journey; a poor understanding of these problems may lead to overly conservative estimates or claims that can not be replicated. Here we argue that researchers can dramatically improve inferences from experiments by focusing on two key issues. First, properly manipulating treatment dispersion, which refers to the variation among levels of a quantitative factor, can improve inference without the need for increasing replicates and sample sizes. Secondly, choosing analytics judiciously, such as selecting between ANOVA and replicated regression for experimental data, can improve inference by contrasting inferential outcomes on the same data. We use language, simple fictional examples, and simulations to show that effect size and power increase with treatment dispersion. We also conducted a small meta-analysis on real data to assess whether the literature confirms in published data that treatment dispersion affects inference. We found that there is no association between treatment dispersion and effect size in published literature, suggesting that some form of bias may be prevalent in published literature. Overall, we found that by focusing on treatment dispersion and analytics, researchers can improve their ability to make sound inferences from their data without the need for increased sample sizes
Applied Artificial Intelligence for Secure and Resilient Smart Grid
No full textThe two-way communication in the power grid enables digitization and automation establishing it as a smart cyber-physical system. However, the integration of cyber layer also makes it vulnerable to security challenges that can compromise grid resilience. The goal of this thesis, therefore, is to advance artificial intelligence (AI)-based models to enhance the security and resilience of the smart grid. This goal is realized through a detailed investigation of notable cyberattack, and the establishment of a sophisticated anomaly prediction model, both enabling stronger defense and protection. As a cyberattack case study, the research focused on a special case of data integrity attack, called false data injection attack (FDIA), due its potentially devastating impacts on grid resilience by misguiding operators with undetected manipulation of measurements. The primary research analyzed the worse case scenario by proposing an FDIA scheme that requires minimized grid topology knowledge and meter accessibility by attackers, using PCA and particle swarm optimization. Next, a multi-objective optimization approach was explored using the SPEA2 algorithm effectively minimizing the number of attacked meters and maximizing the impact, all the while remaining stealthy to demonstrate the trade-off between the objectives. Furthermore, the exploitation of AI by adversaries was investigated by proposing a topology-blind FDIA using only historical measurements for topology inference and attack vector generation. An attacker-side verification of the attack vector using a substitute bad data detect (BDD) ensured reduced chance of detection by the true BDD, revealing enhanced threat. Finally, a sophisticated anomaly prediction model in 5G-based distributed feeder automation was explored. This cross-scenario prediction model combines a mutual information-based FastICA as an anomaly indicator, a dual discriminator conditional GAN for abnormal data prediction, and fuzzy c-Means to classify the predicted data into high, medium and low-risk service degradation due to anomalies. By training on three distinct scenarios, the model predicts future abnormal scenarios based on current normal scenario, facilitating proactive prevention. Overall, this thesis endeavors to make significant contributions to enhancing the security and resilience of smart grids by AI-driven cyberattack investigations, and AI-based anomaly prediction, aiding in defense and prevention mechanisms for a fortified and robust smart grid
System-Level Analysis and Design of Safety-Critical Cyber Physical Systems
Get PDFThe reduction in size and cost of hardware together with the accelerating innovation and advancement in sensor and computational technologies have opened the door for cyber physical systems into all types of applications. While most early systems involved varying degrees of human involvement, the various success stories are encouraging designers to develop cyber physical systems for autonomous control.
The trustworthiness of a cyber-physical system is essential for it to be qualified for utilization in most real-life deployments. This is especially critical for systems that deal with precious human lives. which can be engaged directly as in biomedical systems or indirectly as in automotive systems. Although use-cases for biomedical and automotive systems are considered, the proposed generalized framework can be used to analyze the safety of various cyber-physical systems.
These safety-critical systems can be investigated using both experimental testing and model-based verification. Accurate models have the potential to permit investigating the system behavior under abnormal scenarios. Also, appropriate modeling can speed-up the development process by evaluating candidate designs at an early stage of the design cycle.
Model-based verification can be conducted using the less-exhaustive simulation testing or the resources-greedy model checking. As a trade-off, statistical model checking bears a feasible approach where statistical guarantees can be examined with a specific level of confidence. This research addresses the problem of utilizing accurate system-level models to analyze and design safety-critical cyber-physical systems.
The behavioral descriptions of cyber physical systems are modelled by constructing equivalent formal models. These system-level models are used to conduct statistical model checking to verify properties written using metric interval temporal logic and to provide statistical guarantees on the system safety. This approach is applied on biomedical and automotive systems to verify their safety with consideration for some distortions resulting from unintentional or intentional sources. The proposed verification approach enlightens the development process by providing feedback that can help elect the designs. Moreover, new robust and safe control techniques are proposed to enhance the safety of a closed-loop glucose controller system. Also, a systematic approach is proposed for safety analysis of cyber physical systems. This approach processes systems described using SysML diagrams and applies a new proposed automatic algorithm to construct equivalent formal models. This research work is a step towards bridging the gap between system-level models and formal models so that analysis can be conducted efficiently to enhance the safety and robustness of cyber-physical systems
An Efficient Neural Network Architecture and Training Protocol for 3D Point Cloud Classification
Get PDFThe point cloud is a set of data points in a 3D coordinate system with an irregular data format. As a result, they are needed to be transformed into a collection of images before being fed into models. This unnecessarily increases the volume of the data and increases complexities. The existing literature on point cloud uses a fixed number of points sampled from the whole point cloud as the input. However, with large point cloud data, it is important to consider more points as input to have a better understanding of the scene. The computational expense increases if the input number of points increases for existing networks. Our research contributes to the existing point cloud classification literature in two directions. First, we develop a training protocol for improved point cloud training accuracy on top of the existing PointNet \cite{qi2017pointnet} architecture over the ModelNet10 dataset. A few variations of encoder models have been proposed in this regard. Also, an extensive hyperparameter study and ablation study are done. These experiments achieve a 6.10\% improvement over the baseline model. After that, we propose DualNet, a novel 3D point cloud network that resolves the trade-off between the number of input points and the computational expense of 3D data. The DualNet consists of two branches: DensetNet and SparseNet. The SparseNet is a comparatively large network in terms of number of parameters, that samples a small number of points from the whole point cloud. Whereas the DenseNet is a lightweight network that takes a large number of points as input. SparseNet is composed of more number of channels than DenseNet making it more computationally expensive than DenseNet. While the accuracy of the model shows good improvement when the number of points increases, the overall computational cost of DenseNet does not increase much in such settings. DualNet shows 0.81\% and 0.45\% increase in the SOTA results on ModelNet40 and ScanObjectNN respectively. In respect of computational complexity, our model takes about 40\% less time compared to SOTA
The effects of endogenous and exogenous gonadal hormones on spatial navigation in women and in female rats.
Get PDFGonadal hormones, both endogenous and exogenous, are implicated in cognition. Yet, the role of gonadal hormones in spatial navigation remains relatively unexplored. While navigating a maze either place memory or response memory can be used. The gonadal hormone 17β-estradiol affects which memory system female rats use during navigation, thus producing a memory bias. Across the menstrual cycle, women's memory bias is also altered. This thesis examined the role of endogenous and exogenous hormones in spatial navigation in both female rats and in women. It was shown that the endogenous hormone, progesterone promoted the use of response memory in females. It was also shown that the exogenous hormones used in hormonal contraceptives impact memory bias in females. However, the impact of these hormones on spatial navigation was different in rats than what was observed in humans. It was also shown that both endogenous and exogenous hormones have different effects depending on whether they are administered alone or in combination. Both endogenous and exogenous gonadal hormones are involved in memory bias during spatial navigation in females
Ultra-Dense Networks in 5G and Beyond: Challenges and Promising Solutions
Get PDFUltra-Dense Network (UDN) is one of the promising and leading directions in Fifth Generation and beyond (5GB) networks. In UDNs, Small Cells (SCs) or Small Base Stations (SBSs) such as microcells, picocells, or femtocells are deployed in high densities where inter-site distances are within the range of few or tens of meters. UDNs also require that SCs are typically deployed in relatively large densities compared to the Human-Type Communication Users (HTCUs) such as smartphones, tablets, and/or laptops. Such SCs are characterized by their low transmission powers, small coverage areas, and low cost. Hence, the deployment of the SCs can be done either by the cellular network operators or by the customers themselves within their premises to maintain certain levels of Quality of Service (QoS).
However, the randomness of the deployment of the SCs along with the small inter-site distances may degrade the achievable performance due to the uncontrolled Inter-Cell Interference (ICI). Therefore, idle mode capability is an inevitable feature in the high-density regime of SCs. In idle mode, a SC is switched off to prevent ICI when no user is associated to it. In doing so, we can imagine the UDN as a mobile network that keeps following the users to remain as close as possible to them.
In 5G, different use cases are required to be supported such as enhanced Mobile Broad-Band (eMBB), Ultra-Reliable and Low-Latency Communication (URLLC), and massive Machine-Type Communication (mMTC). On one hand, the inevitable upcoming era of smart living requires unprecedented advances in enabling technologies to support the main building blocks of this era which are Internet of Things (IoT) devices. Machine-Type Communication (MTC), the cellular version of Machine-to-Machine (M2M) communication, constitutes the main enabling technology to support communications among such devices with minimal or even without human intervention. The massive number of these devices, Machine-Type Communication Devices (MTCDs), and the immense amount of traffic generated by them require a paramount shift in cellular and non-cellular wireless technologies to achieve the required connectivity. On the other hand, the sky-rocketing number of data hungry applications installed on human-held devices, or HTCUs, such as video conferencing and virtual reality applications require their own advances in the wireless infrastructure in terms of high capacity, enhanced reliability, and reduced latency.
Throughout this thesis, we exploit the UDN infrastructure integrated with other 5G resources and enabling technologies to explore the possible opportunities in supporting both HTC and MTC, either solely or simultaneously. Given the shorter distances between transmitters and receivers encountered in UDNs, more realistic models of the path loss must be adopted such as the Stretched Exponential Path Loss (SEPL) model. We use tools from stochastic geometry to formulate novel mathematical frameworks that can be used to investigate the achievable performance without having to rely on extensive time-consuming Monte-Carlo simulations. Besides, the derived analytical expressions can be used to tune some system parameters or to propose some approaches/techniques that can be followed to optimize the performance of the system under certain circumstances.
Tackling practical scenarios, the complexity, or sometimes in-feasibility, of providing unlimited backhaul capacity for the massive number of SCs must be considered. In this regard, we adopt multiple-association where each HTCU is allowed to associate with multiple SCs. By doing so, we carefully split the targeted traffic among several backhaul links to mitigate the bottleneck forced by limited backhaul capacities. It is noteworthy that for coexisting MTCDs with the HTCUs, activating more SCs would allow more MTCDs to be supported without introducing additional ICI towards the HTCUs.
Targeting different application, multiple-association can be also adopted to tackle computation-intensive applications of HTCUs. In particular, for applications such as augmented reality and environment recognition that require heavy computations, a task is split and partially offloaded to multiple SCs with integrated Edge Computing Servers (ECSs). Then, the task partitions are processed in parallel to reduce the end-to-end processing delay. Based on relative densities between HTCUs and SCs, we use tools from stochastic geometry to develop an offline adaptive task division technique that further reduces the average end-to-end processing delay per user.
With the frequent serious data breaches experienced in recent years, securing data has become more of a business risk rather than an information technology (IT) issue. Hence, we exploit the dense number of SCs found in UDN along with Physical Layer Security (PLS) protocols to secure data transfer. In particular, we again adopt multiple-association and split the data of HTCUs into multiple streams originating from different SCs to prevent illegitimate receivers from eavesdropping.
To support massive number of MTCDs, we deploy the Non-Orthogonal Multiple-Access (NOMA) technique. Using power NOMA, more than one device can be supported over the same frequency/time resource and their signals are distinguished at the receiver using Successive Interference Cancellation (SIC). In the same scope, exploiting the available resources in 5G and beyond networks, we investigate a mMTC scenario in an UDN operating in the Millimeter Wave (mmWave) band and supported by wireless backhauling. In doing so, we shed lights on the possible gains of utilizing the mmWave band where the severe penetration losses of mmWave can be exploited to mitigate the significant ICI in UDNs. Also, the vast bandwidth available in the mmWave band helps to allocate more Resource Blocks (RBs) per SCs which corresponds to supporting more MTCDs
Design of a High-Speed, High-Power Density Permanent Magnet Synchronous Motor for Electric Vehicle Traction
No full textA high-speed electric motor is developed as part of a high-density traction system that includes the motor and the power electronics stage. To consider the close interdependence of the inverter and the motor sizing for a specific application, the high-level design of the high-density system and the detailed design of the electric motor are presented in this thesis.
The main objective of the research work is to improve the state-of-the art power density of a traction system for a passenger car. At the system level, different electronic topologies are discussed, and different types of permanent magnets motor topologies are presented and compared regarding the objective of power density. A modeling methodology is presented as an accurate and flexible tool to evaluate the electromagnetic performances of different PMSM topologies.
One development axis to increase the power density is an enhanced cooling design of the motor heat extraction system. The proposed motor cooling design allows a distributed heat extraction, with a focus on the typically limiting hot spots on the end windings and the rotor magnets.
A second development axis to consider for the motor power density is to increase its speed. A PMSM rotor topology with a retaining sleeve is selected as the high-speed rotor design. The geometric variables that impact the electromagnetic performance and losses are evaluated with some considerations on the demagnetization vulnerability. The thermal management of the losses is discussed.
A comparison of the copper losses for different wire types and slot shapes is presented, with a focus on the high frequency losses. A stator slot configuration is proposed as a good compromise between high frequency losses reduction, DC copper resistance, heat transfer, performance, and cost.
A high-density system design is presented, combining the different motor enhancements proposed in the research work. From the system geometry and simulated output power, a power density exceeding the targeted value is achieved. The performances are compared with those of a recent design from the industry. Over the full speed range, both peak and continuous power density are higher for the proposed high-density system