Search results
(2,861 - 2,880 of 2,980)
Pages
- Title
- Deep Learning Methods For Wireless Networks Optimization
- Creator
- Zhang, Shuai
- Date
- 2022
- Description
-
The resurgence of deep learning techniques has brought forth fundamental changes to how hard problems could be solved. It used to be held that...
Show moreThe resurgence of deep learning techniques has brought forth fundamental changes to how hard problems could be solved. It used to be held that the solutions to complex wireless network problems require accurate mathematical modeling of the network operation, but now the success of deep learning has shown that a data-driven method could generate powerful and useful representations such that the problem could be solved efficiently with surprisingly competent performance. Network researchers have recognized this and started to capitalize on the learning methods’ prowess. But most works follow the existing black-box learning paradigms without much accommodation to the nature and essence of the underlying network problems. This thesis focuses on a particular type of classical problem: multiple commodity flow scheduling in an interference-limited environment. Though it does not permit efficient exact algorithms due to its NP-hard complexity, we use it as an entry point to demonstrate from three angles how the learning-based methods can help improve the network performance. In the first part, we leverage the graphical neural network (GNN) techniques and propose a two-stage topology-aware machine learning framework, which trains a graph embedding unit and a link usage prediction module jointly to discover links that are likely to be used in optimal scheduling. The second part of the thesis is an attempt to find a learning method that has a closer algorithmic affinity to the traditional DCG method. We make use of reinforcement learning to incrementally generate a better partial solution such that a high quality solution may be found in a more efficient manner. As the third part of the research, we revisit the MCF problem from a novel viewpoint: instead of leaning on the neural networks to directly generate the good solutions, we use them to associate the current problem instance with historical ones that are similar in structure. These matched instances’ solutions offer a highly useful starting point to allow efficient discovery of the new instance’s solution.
Show less
- Title
- Essays on Clean Energy Finance and Cryptocurrency Market
- Creator
- Xie, Yao
- Date
- 2021
- Description
-
This dissertation includes four essays with several empirical investigations in the areas of clean energy finance and cryptocurrencies.In the...
Show moreThis dissertation includes four essays with several empirical investigations in the areas of clean energy finance and cryptocurrencies.In the first essay, I investigate the heterogeneous relationship between various determinants of the clean energy market across all subsectors of the clean energy stock market. My findings reveal that VIX is the most significant predictor of all clean energy subsectors conditional volatility. During the COVID-19 stress period, economic uncertainty measures become more significant measures. The heterogeneity of clean energy market persists in the out-of-sample results. These results suggest that portfolio diversification for different clean energy subsector is necessary. In the second essay, I study the safe haven property of several volatility indexes on clean energy subsectors. I compare the current COVID-19 stress period and the time before. The results show that market volatility and commodity volatility are good safe haven assets during the COVID-19 period. But they are not safe haven assets against the clean energy subsector before the pandemic period. Among all volatility indexes, gold volatility index is the most effective safe haven assets. In the third essay, I investigate the characteristics of Bitcoin as a financial asset. A comprehensive set of information variables under five categories: macroeconomics, blockchain technology, other markets, stress level, and investor sentiment. The empirical results show that blockchain technology, stress level and investor sentiment have strong predicting power on Bitcoin returns. In the fourth essay, I aim to study how extreme sentiment measures from Google Trend and Wikipedia Pageviews affect both traditional cryptocurrency, such as Bitcoin and stablecoin, like Tether. Our results show that Tether’s return is not affected by the extreme sentiment measures during the COVID-19 stress period which suggests that stablecoin can offer price stability.
Show less
- Title
- DEEP LEARNING AND COMPUTER VISION FOR INDUSTRIAL APPLICATIONS: CELLULAR MICROSCOPIC IMAGE ANALYSIS AND ULTRASOUND NONDESTRUCTIVE TESTING
- Creator
- Yuan, Yu
- Date
- 2022
- Description
-
For decades, researchers have sought to develop artificial intelligence (AI) systems that can help human beings on decision making, data...
Show moreFor decades, researchers have sought to develop artificial intelligence (AI) systems that can help human beings on decision making, data analysis and pattern recognition applications where analytical methods are ineffective. In recent years, Deep Learning (DL) has been proven to be an effective AI technique that can outperform other methods in applications such as computer vision, natural language processing, autonomous driving. Realizing the potential of deep learning techniques, researchers have also started to apply deep learning on other industrial applications. Today, deep learning based models are used to innovate and accelerate automation, guidance, and decision making in various industries including automotive industry, pharmaceutical industry, finance, agriculture and more. In this research, several important industrial applications (on Biomedicine and Non-Destructive Testing) utilizing deep learning algorithms will be introduced and analyzed. The first biopharmaceutical application focuses on developing a deep learning based model to automate the visual inspection process in Median Tissue Culture Infectious Dose(TCID50). TCID50 is one of the most popular methods for viral quantification. An important step of TCID50 is to visually inspect the sample and decide if it exhibits cytopathic effect(CPE) or not. Two novel models have been developed to detect CPE in microscopic images of cell culture in 96 well-plates. The first model consists of a convolutional neural network (CNN) and support vector machine(SVM). The second model is a fully convolutional network (FCN) followed by morphological post-processing steps. The models are tested on 4 cell lines and achieve very high accuracy. Another biopharmaceutical application developed for cellular microscopic images is the clonal selection. Clonal selection is one of the mandatory steps in cell line development process. It focuses on verifying the clonality of the cell culture. The researchers used to visually inspect the microscopic images to verify the clonality. In this work, a novel deep learning based model and a workflow is developed to accelerate the process. This algorithm consists of multiple steps, including image analysis after incubation to detect the cell colonies, and verify its clonality in day0 image. The results and common mis-classification cases are shown in this thesis. Image analysis method is not the only technology that has been advancing for cellular image analysis in biopharmaceutical industry. A new class of instruments are currently used in biopharmaceutical industry which enable more opportunities for image analysis. To make the most of these new instruments, a convolutional neural network based architecture is used to perform accurate cell counting and cell morphology based segmentation. This analysis can provide more insight of the cells at very early stage in characterization process of cell line development. The architecture and the testing results are presented in this work. The proposed algorithm has achieved very high accuracy on both applications, and the cell morphology based segmentation enables a brand new feature for scientists to predict the potential productivity of the cells. Next part of this dissertation is focused on hardware implementation of Ultrasonic Non-Destructive Testing (NDT) methods based on deep learning, which can be highly useful in flaw detection and classification applications. With the help of a smart and mobile Non-Destructive Testing device, engineers can accurately detect and locate the flaws inside the materials without reliance on high performance computation resources. The first NDT application presents a hardware implementation of a deep learning algorithm on Field-programmable gate array(FPGA) for Ultrasound flaw detection. The Ultrasound flaw detection algorithm consists of a wavelet transform followed by a LeNet inspired convolutional neural network called Ultra-LeNet. This work is focused on implementing the computationally difficult part of this algorithm: Ultra-LeNet, so that it can be used in the field where high performance computation resources (e.g., AWS) are not accessible. The implementation uses resource partitioning to design two dedicated pipelined accelerators for convolutional layers and fully connected layers respectively. Both accelerators utilize loop unrolling, loop pipelining and batch processing techniques to maximize the throughput. The comparison to other work has shown that the implementation has achieved higher hardware utilization efficiency. The second NDT application is also focused on implementing a deep learning based algorithm for Ultrasound flaw detection on a FPGA. Instead of implementing the Ultra-LeNet, the deep learning model used in this application is Meta-learning based Siamese Network, which is capable for multi-class classification and it can also classify a new class even if it does not appear in the training dataset with the help of automated learning features. The hardware implementation is significantly different than the previous algorithm. In order to improve the inference operation efficiency, the model is compressed with both pruning and quantization, and the FPGA implementation is specifically designed to accelerate the compressed CNN with high efficiency. The CNN model compression method and hardware design are novel methods introduced in this work. Comparison against other compressed CNN accelerators is also presented.
Show less
- Title
- MARKETABLE LIMIT ORDERS AND NON-MARKETABLE LIMIT ORDERS ON NASDAQ
- Creator
- ZHANG, DAN
- Date
- 2022
- Description
-
My research includes two parts. In the first part of my research, I classify marketable limit orders into three different types: large...
Show moreMy research includes two parts. In the first part of my research, I classify marketable limit orders into three different types: large marketable order to buy, large marketable order to sell, and small marketable order. I use dummy variance method to research the effect of the three marketable orders on standardized variance, and find that LMOB and LMOS play significant role in variance increase. The second part of my research is about modelling of time to execution and time to cancellation of Non-marketable limit orders. I construct variables and model time to execution for NLO to buy and time to cancellation for NLO to buy and NLO to sell based on exponential distribution with accelerated failure time specification. My research shows that the longer the distance of limit price to buy away from the best bid price, the longer time to execution is. The longer the distance of limit price to buy away from the best bid price or limit price to sell away from the best ask price, the longer the time to cancellation is.
Show less
- Title
- Stochastic dynamical systems with non-Gaussian and singular noises
- Creator
- Zhang, Qi
- Date
- 2022
- Description
-
In order to describe stochastic fluctuations or random potentials arising from science and engineering, non-Gaussian or singular noises are...
Show moreIn order to describe stochastic fluctuations or random potentials arising from science and engineering, non-Gaussian or singular noises are introduced in stochastic dynamical systems. In this thesis we investigate stochastic differential equations with non-Gaussian Lévy noise, and the singular two-dimensional Anderson model equation with spatial white noise potential. This thesis consists of the following three main parts. In the first part, we establish a linear response theory for stochastic differential equations driven by an α-stable Lévy noise (1<α<2). We first prove the ergodic property of the stochastic differential equation and the regularity of the corresponding stationary Fokker-Planck equation. Then we establish the linear response theory. This result is a general fluctuation-dissipation relation between the response of the system to the external perturbations and the Lévy type fluctuations at a steady state.In the second part, we study the global well-posedness of the singular nonlinear parabolic Anderson model equation on a two-dimensional torus. This equation can be viewed as the nonlinear heat equation with a random potential. The method is based on paracontrolled distribution and renormalization. After splitting the original nonlinear parabolic Anderson model equation into two simpler equations, we prove the global existence by some a priori estimates and smooth approximations. Furthermore, we prove the uniqueness of the solution by classical energy estimates. This work improves the local well-posedness results in earlier works.In the third part, we investigate the variation problem associated with the elliptic Anderson model equation in a two-dimensional torus in the paracontrolled distribution framework. The energy functional in this variation problem is arising from the Anderson localization. We obtain the existence of minimizers by a direct method in the calculus of variations, and show that the Euler-Lagrange equation of the energy functional is an elliptic singular stochastic partial differential equation with the Anderson Hamiltonian. We further establish the L2 estimates and Schauder estimates for the minimizer as weak solution of the elliptic singular stochastic partial differential equation. Therefore, we uncover the natural connection between the variation problem and the singular stochastic partial differential equation in the paracontrolled distribution framework.Finally, we summarize our results and outline some research topics for future investigation.
Show less
- Title
- Expanding the Magic Circle and the Self: Integrating Discursive Topics into Games
- Creator
- da Rosa Faller, Roberto
- Date
- 2020
- Description
-
This study focuses on games for self-development and how they communicate ideas, challenge established assumptions, cause reflection, and...
Show moreThis study focuses on games for self-development and how they communicate ideas, challenge established assumptions, cause reflection, and provoke change. It explores the integration of discursive topics – specifically those perceived as difficult, political, philosophical, taboo, or controversial – into games, and how to manage player exposure to these topics through design while avoiding player disengagement to achieve self-development goals. Using a Research Through Design approach, this study was conducted in two phases. The first exploratory phase resulted in an analytical framework with four distinct lenses: engaging play experience; player’s emotional investment; the friction points of discursive topics; and, controlled exposure to the topic. During the second phase, this framework was used to analyze eight case studies and three prototypes. The resultant insights from analysis revealed five categories – topic depiction, emotional climate, emotional anchors, topic delivery, and exposure timing – that form the Discursive Topic Integration Framework for self-development. This framework offers a new theoretical perspective for design scholars and practicing designers about how to manipulate the “magic circle” (a safe temporary space for the act of play), by intentionally designing for discursive topics and their friction points. It contributes strategies about when, how, how frequently, and with what intensity discursive topics may be introduced and abstracted in games. It frames the discursive topic, creates the emotional climate, and anchors the player inside the magic circle of the game so that they feel engaged, motivated, and curious without becoming overwhelmed. This study also generated two additional frameworks, including: the Self-Development Opportunity Matrix that can be used to generate or evaluate self-development goals; and, the Five Categories of Transitional and Traumatic Experiences that can assist in the design of games and other experiences that build a person’s capacity, self-determination, and commitment to positive change.
Show less
- Title
- Development of a novel ultra-nanocrystalline diamond (UNCD) based photocathode and exploration of its emission mechanisms
- Creator
- Chen, Gongxiaohui
- Date
- 2020
- Description
-
High quality electron sources are one of the most commonly used probing tools used for the study of materials. Photoemission cathodes, capable...
Show moreHigh quality electron sources are one of the most commonly used probing tools used for the study of materials. Photoemission cathodes, capable of producing ultra-short and ultra-high intensity beams, are a key component of accelerator based light sources and some microscopy tools. High quantum efficiency (QE), low intrinsic emittance, and long lifetime (or good vacuum tolerance) are three of the most critical features for a photocathode; however, these are difficult to achieve simultaneously and trade-offs need to be made for different applications. In this work, a novel semi-metallic material of nitrogen-incorporated ultrananocrystalline diamond ((N)UNCD) has been studied as a photocathode. (N)UNCD has many of the unique diamond properties, such as low intrinsic as-grown surface roughness (at the order of 10~nm) due to its nanometer scale crystalline size, relatively long lifetime in air, high electrical conductivity with nitrogen doping, and potentially high QE performance due to the high grain boundary densities where most of electron emission occurs. High contrast interference of incident and reflected radiation within (N)UNCD thin films was observed, and this feature allows fast thickness determination based on an analytical optics methodology. This method has been extended to study and calculate the etching rates of two commonly used O$_2$ and H$_2$ plasmas for use with future (N)UNCD microfabrication processes. The mean transverse energy (MTE) of (N)UNCD was determined over a wide UV range in a DC photogun. Unique MTE behavior was observed; it did not scale with photon energy unlike most metals. This behavior is associated with emission from spatially-confined states in the graphite regions (with low electron effective mass) between the diamond grains. Such behavior suggests that beam brightness many be increased by the simple mechanism of increasing the photon energy so that the QE increases, while the MTE remains constant.Two individual (N)UNCD photocathodes synthesized two years apart have been characterized in a realistic RF photogun. Both the QE and intrinsic emittance were characterized. It was found that the QE of $\sim4.0\times 10^{-4}$, is more than an order of magnitude higher than that of most commonly used metal cathodes (such as Cu and Nb). The intrinsic emittance (0.997~$\mu$m/mm) is comparable to that of photocathodes now deployed in research accelerators. The most impressive feature is the excellent robustness of (N)UNCD material; there was no evidence of performance degradation, even after years-long atmospheric exposure. The results of this work demonstrate that a cathode made of (N)UNCD material is able to achieve balanced performance of three of the primary critical photocathode figures-of-merit.
Show less
- Title
- WASTEWATER COLLECTION SYSTEM MODELING: TOWARDS AN INTEGRATED URBAN WATER AND ENERGY NETWORK
- Creator
- Wang, Xiaolong
- Date
- 2020
- Description
-
Wastewater collection systems, among the oldest features of urban infrastructure, are typically dedicated to collect and transport wastewater...
Show moreWastewater collection systems, among the oldest features of urban infrastructure, are typically dedicated to collect and transport wastewater from users to water resource recovery facilities (WRRFs). Since the 1970s, wastewater engineers and scientists have come to understand that wastewater collection systems can bring benefits for urban water and energy networks, including thermal energy recovery and converting pipelines to bioreactors. However, there is little knowledge about the temporal and spatial changes of collection systems parameters that are important for these applications. Furthermore, the vast majority of existing studies of these applications have focused on laboratory or extremely small-scale systems; there have been few studies about beneficial applications associated with large-scale systems. The purpose of this study is to increase our understanding of how urban wastewater collection systems can bring potential benefits to urban water and energy systems. Models describing wastewater hydraulics, temperature, and water quality can provide valuable information to help evaluate thermal energy recovery and wastewater pretreatment feasibility. These kinds of models, and supporting data from a case study, were used in this study; sizes of the theoretical wastewater collection systems range from 2.6 L/s to 52 L/s, and the sample locations of the case study had flows ranging from 2.3 L/s to 24.5 L/s. A cost-benefit analysis of wastewater source heat pumps was used to evaluate the thermal energy recovery feasibility for different sizes of wastewater collection systems. Results show that the large collection system can support a large capacity heat pump system with a relatively low unit initial cost. Small collection systems have a slightly lower unit operating cost due to the relatively high wastewater temperature. When the heat pump system capacity design was based on the average available energy from the collection system, larger systems have lower payback times; the lowest payback time is about 3.5 years. The wastewater quality model was used to describe the dissolved oxygen (DO) and organic matter concentrations changes in the collection system. The model provides a framework for predicting pretreatment capability. Model results show that DO concentration is the limiting parameter for organic matter removal. Larger collection systems can provide more organic matter removal because they provide relatively longer retention times, and they offer the potential for greater DO reaeration. The model can also be used to identify environmental conditions in sewer pipelines, providing information for potential issues predication.
Show less
- Title
- COMPREHENSIVE ANALYSIS OF EXON SKIPPING EDITS WITHIN DYSTROPHIN D20:24 REGIONS
- Creator
- Niu, Xin
- Date
- 2020
- Description
-
Exon skipping is a disease modifying therapy that operates at the RNA level. In this strategy, oligonucleotide analog drugs are used to...
Show moreExon skipping is a disease modifying therapy that operates at the RNA level. In this strategy, oligonucleotide analog drugs are used to specifically mask specific exons and prevent them from being included in the mature mRNA. Exon skipping can also be used to restore protein expression in cases where a genetic frameshift mutation has occurred, and this how it is applied to Duchenne muscular dystrophy, DMD. DMD most commonly arises as a result of large exonic deletions that juxtapose flanking exons of incompatible reading frame, which abolishes dystrophin protein expression. This loss leads to the pathology of the disease, which is severe, causing death generally in the second or third decade of life. Here, the primary aim of exon skipping is to restore the reading frame by skipping an exon adjacent to the patient’s original. While restoring some protein expression is good, how removing some region from the middle of protein affects its structure and function is unclear. Complicating this in this case is that the dystrophin gene is very large, containing 79 exons. Many different underlying deletions are knowns, and exon skipping can be applied in many ways. It has previously been shown that many exon-skip edits result in structural perturbations of varying degrees. Very few studies are focused on the protein biophysical study and it is still basically unclear whether and how such editing can be done to minimize such perturbations. In order to provide the solid evidences which prove the significant variation among those cases (especially for the clinically relevant cases) and better understanding the general principles of “what makes a good edit”, we examine a systematic and comprehensive panel of possible exon edits in a region of the dystrophin protein. The domain D20:24 of dystrophin rod region are selected for its entirety which is separated by hinge region (mostly random coiled structure) and addition of other STRs will not disrupt the structure stability. Also D20:24 regions lie in the Hot Spot region II (HS2) which holds the most number of DMD patients. During the comprehensive scan, we identify for the first time, exon edits that appear to maintain structural stability similar to wild-type protein and those clinically relevant edits. Then we figure out the factors that appear to be correlated with the degree of structural perturbation, such as the number of cooperative protein domains, as well as how the edited exon structure interacts with the protein domain structure. Our study is the first systematic and comprehensive scan for an entire multiple STRs domain. This would help us understand the protein nature of various exon skipping edits and provide useful target for clinical treatment. Also the knowledge we learned may be applied to produce more sophisticated CRISPR edits in the future work.
Show less
- Title
- DEVELOPMENT OF FULLY BIOCOMPATIBLE HYDROGEL NANOPARTICLE FORMULATIONS FOR CONTROLLED-RELEASE DELIVERY OF A WIDE VARIETY OF BIOMOLECULES
- Creator
- Borges, Fernando Tancredo Pereira
- Date
- 2020
- Description
-
In recent years, our group has focused on the production of PEGDA-based hydrogel scaffolds and nanoparticles for drug delivery of small...
Show moreIn recent years, our group has focused on the production of PEGDA-based hydrogel scaffolds and nanoparticles for drug delivery of small molecules. However, with recent advances in modern therapeutic treatments, such as protein and genetic engineering, there is an increasing need for the development of drug delivery devices that would be able encapsulate larger molecules. Therefore, the goal of this thesis work was to develop a systematic way to produce fully biocompatible PEGDA-based hydrogel nanoparticle formulations that would be able to encapsulate any size molecule, ranging from small ionic molecules, to peptides and proteins, all the way to large nucleic acids, and deliver it in a controlled manner.The first of part of this work consisted of developing a stable and reproducible process for the production of hydrogel PPi-NPs. Initial studies were done in order to assess the influence of phosphate salts in the polymerization system and it was found that both monophosphate and polyphosphate salts significantly damper the NVP homo-polymerization kinetics, but do not affect the co-polymerization of NVP and PEGDA. Then, emulsion stability studies were done to determine whether phosphate salts affected the stability of the minimeulsion system used in the production of the nanoparticles. Cloud point measurements and droplet size screening measurements showed that by transitioning from a Pi-loaded emulsion system to a PPi-loaded emulsion system, the required HLB of the emulsion shifts by 1.5 points. Upon correction for that shift, a reproducible process for production of PPi-loaded nanoparticles was obtained. A parametric study was then performed to see how the different process parameters affected the different properties of the produced particles. The second part of the work consisted in developing a platform for encapsulation of large to very-large molecules within these hydrogel systems. A new set of equations was developed for better estimation of the interstitial space, available for encapsulation of molecules, of crosslinked polymers that used very high molecular weight crosslinkers and/or high amounts of crosslinker. Upon development of this new set of equations, hydrogel discs were made via photopolymerization in order to validate the equations. By introducing a third monomer, EGA, and varying the molecular weight and concentration of the crosslinker, hydrogels with a wide range of mesh dimensions from 25 to 700 were achieved. These gels were then used to encapsulate 4 different sample molecules of varying molecular weights and size. A new heuristic was developed for encapsulation of non-spherical molecules, where the aspect ratios of the molecule and of the polymer network are considered. By varying the size of the ratios of the dimensions of the hydrogel network to the dimensions of the molecule, significantly different release profiles of small molecules, peptides and oligonucleotides were obtained. Finally, in order to explore different administration routes, the process was transitioning into being fully biocompatible. The organic solvent previously used in the emulsion system was replaced by soybean oil and the surfactants were replaced by a food-grade surfactant, PGPR, to form Bio-Compatible Nanoparticle Emulsions (BCNEs). Qualitative release from the BCNEs was shown. A new method for quantitative measuring of release from BCNE was developed. Release from QK-BCNE was observed up to 46 days, which is unprecedented for sustained-release and revolutionary for the field. A BCNE spreadable ointment formulation was also developed.
Show less
- Title
- Development of Human Brain Atlas Resources
- Creator
- Qi, Xiaoxiao
- Date
- 2020
- Description
-
Digital human brain atlases play an increasingly critical role and are widely used in neuroimaging studies such as developing biomarkers,...
Show moreDigital human brain atlases play an increasingly critical role and are widely used in neuroimaging studies such as developing biomarkers, training data for machine learning algorithms, functional connectivity analysis and so on. A brain atlas typically consists of brain templates of different imaging modalities that are representative of individual brains under study in a standard atlas space and semantic labels that delineate brain regions according to the characteristics of the underlying tissue.The IIT Human Brain Atlas project has developed the state-of-the-art diffusion tensor imaging (DTI) template, high angular resolution diffusion imaging (HARDI) template, and anatomical templates for the young adult brain in a standardized space. The probabilistic maps of gray matter (GM) labels and tissue segmentations were also constructed based on the anatomical information of the atlas. This thesis introduced an enhanced T1-weighted template that were developed by combining information from both diffusion and anatomical data. The GM labels and tissue segmentation maps in the standardized space were also improved. Existing white matter (WM) atlases typically lack specificity in terms of brain connectivity. A new approach named regionconnect was developed in this work based on precalculated average healthy adult brain connectivity information stored in standard space in a fashion that allows fast retrieval and integration. This thesis first generated and evaluated the white matter connectome of the IIT Human Brain Atlas v.5.0. Next, the new white matter connectome was used to develop multi-layer, connectivity-based labels for each white matter voxel of the atlas, consistent with the fact that each voxel may contain axons from multiple connections. The regionconnect algorithm was then developed to rapidly integrate information contained in the multi-layer labels across voxels of a white matter region and to generate a list of the most probable connections traversing that region. The regionconnect algorithm as well as the white matter tractogram and connectome, multi-layer, connectivity-based labels, and associated resources developed for the IIT Human Brain Atlas v.5.0 in this work are available at www.nitrc.org/projects/iit. Furthermore, it was well established that use of a young adult atlas in studies of older adults is inappropriate due to the age-related characteristic changes of the brain, resulting in an increasing demand of digital brain atlases for the older adults. To fulfill this demand, a function of fiber orientation distribution (fODF) template that is representative of older adults was developed in a standardized atlas space for studies of white matter of older adult human brains, which built a solid foundation for the development of the white matter resources for the older adults human brain atlas.
Show less
- Title
- Development of Microfluidic Platform to Study Insulin Resistance
- Creator
- Tanataweethum, Nida
- Date
- 2020
- Description
-
Insulin resistance, a precursor for the development of type 2 diabetes (T2D), propagates among heterologous tissues through dysregulated lipid...
Show moreInsulin resistance, a precursor for the development of type 2 diabetes (T2D), propagates among heterologous tissues through dysregulated lipid flux, as well as dysregulated glucose production, and secretion of cytokines, adipokines and hepatokines. Although T2D is characterized by systemic insulin resistance, disruption of insulin signaling in the liver and adipose tissue recapitulates many aspects of T2D, including enhance endogenous glucose production as well as defects of insulin action. Mechanistic studies often aim to provide fundamental understanding of the observations from human and animal studies. Due to the complexity of animal models and the multifactorial character of T2D, there is a strong need to develop advanced experimental systems such as in vitro models that can enable the recapitulation of the complex physiology of the in vivo system and enable investigation of the pathological pathways as well as identify novel treatment options. The overall goal of this study was to develop insulin resistant models of adipose tissue and liver to study the metabolic function of each organ as well as to the organ-organ crosstalk. To accomplish this goal, four specific aims were pursued: (1) Establish adipose tissue on-a-chip to study the metabolic function of the adipocytes in flow culture; (2) Develop towards an insulin resistant adipose on-a-chip to study the metabolic function of adipocytes in setting of insulin resistance; (3) Develop insulin resistant liver on-a-chip to investigate the metabolic function of hepatocytes in setting of insulin resistance; (4) Develop adipose-liver on-a-chip in setting of insulin resistance to identify the metabolic interaction between organs.
Show less
- Title
- Exploiting contextual information for deep learning based object detection
- Creator
- Zhang, Chen
- Date
- 2020
- Description
-
Object detection has long been an important research topic in computer vision area. It forms the basis of many applications. Despite the great...
Show moreObject detection has long been an important research topic in computer vision area. It forms the basis of many applications. Despite the great progress made in recent years, object detection is still a challenging task. One of the keys to improving the performance of object detection is to utilize the contextual information from the image itself or from a video sequence. Contextual information is defined as the interrelated condition in which something exists or occurs. In object detection, such interrelated condition can be related background/surroundings, support from image segmentation task, and the existence of the object in the temporal domain for video-based object detection. In this thesis, we propose multiple methods to exploit contextual information to improve the performance of object detection from images and videos.First, we focus on exploiting spatial contextual information in still-image based object detection, where each image is treated independently. Our research focuses on extracting contextual information using different approaches, which includes recurrent convolutional layer with feature concatenation (RCL-FC), 3-D recurrent neural networks (3-D RNN), and location-aware deformable convolution. Second, we focus on exploiting pixel-level contextual information from a related computer vision task, namely image segmentation. Our research focuses on applying a weakly-supervised auxiliary multi-label segmentation network to improve the performance of object detection without increasing the inference time. Finally, we focus on video object detection, where the temporal contextual information between video frames are exploited. Our first research involves modeling short-term temporal contextual information using optical flow and modeling long-term temporal contextual information using convLSTM. Another research focuses on creating a two-path convLSTM pyramid to handle multi-scale temporal contextual information for dealing with the change in object's scale. Our last work is the event-aware convLSTM that forces convLSTM to learn about the event that causes the performance to drop in a video sequence.
Show less
- Title
- ENLARGED PERIVASCULAR SPACES IN COMMUNITY-BASED OLDER ADULTS
- Creator
- Javierre Petit, Carles
- Date
- 2020
- Description
-
Enlarged perivascular spaces (EPVS) have been associated with aging, increased stroke risk, decreased cognitive function and vascular dementia...
Show moreEnlarged perivascular spaces (EPVS) have been associated with aging, increased stroke risk, decreased cognitive function and vascular dementia. Furthermore, recent studies have investigated the links of EPVS with the glymphatic system (GS), since perivascular spaces are thought to play a major role as the main channels for clearance of interstitial solutes from the brain. However, the relationship of EPVS with age-related neuropathologies is not well understood. Therefore, more conclusive studies are needed to elucidate specific relationships between EPVS and neuropathologies. After demonstration of their neuropathologic correlates, detailed assessment of EPVS severity could provide as a potential biomarker for specific neuropathologies.In this dissertation, our focus was twofold: to develop a fully automatic EPVS segmentation model via deep learning with a set of guidelines for model optimization, and to evaluate both manual and automatic assessment of EPVS severity to investigate the neuropathologic correlates of EPVS, and their contribution to cognitive decline, by combining ex-vivo brain magnetic resonance imaging (MRI) and pathology (from autopsy) in a large community-based cohort of older adults. This project was structured as follows. First, a manual approach was used to assess neuropathologic and cognitive correlates of EPVS burden in a large dataset of community-dwelling older adults. MR images from each participant were rated using a semiquantitative 4-level rating scale, and a group of identified EPVS was histologically evaluated. Two groups of participants in descending order of average cognitive impairment were defined based and studied. Elasticnet regularized ordinal logistic regression was used to assess the neuropathologic correlates of EPVS burden in each group, and linear mixed effects models were used to investigate the associations of EPVS burden with cognitive decline. Second, a fully automatic EPVS segmentation model was implemented via deep learning (DL) using a small dataset of 10 manually segmented brain MR images. Multiple techniques were evaluated to optimize performance, mainly by implementing strategies to reduce model overfitting. The final segmentation model was evaluated in an independent test set and the performance was validated with an expert radiologist. Third, the DL segmentation model was used to segment and quantify EPVS. Quantified EPVS (qEPVS) were evaluated by combining ex-vivo MRI, pathology, and longitudinal cognitive evaluation. EPVS quantification allowed to study qEPVS both in the whole brain and regionally. Two different qEPVS metrics were studied. Elastic-net regularized linear regression was used to assess the neuropathologic correlates of qEPVS within each region of interest (ROI) under study, and linear mixed effects models were used to investigate the associations of qEPVS with cognitive decline. Finally, a preliminary study investigated the longitudinal associations of qEPVS with time. The DL segmentation model was re-trained using 4 in-vivo MR images. EPVS were segmented and quantified in a large longitudinal cohort where each participant was imaged at multiple timepoints. Factors that influenced segmentation performance across timepoints were evaluated, and linear mixed effects models controlling for these factors were used to investigate the associations of qEPVS with time.
Show less
- Title
- BIG DATA AS A SERVICE WITH PRIVACY AND SECURITY
- Creator
- Hou, Jiahui
- Date
- 2020
- Description
-
With the increase of data production sources like IoT devices (e.g., smartwatches, smartphones) and data from smart home (health sensor,...
Show moreWith the increase of data production sources like IoT devices (e.g., smartwatches, smartphones) and data from smart home (health sensor, energy sensors), truly mind-boggling amounts of data are generated daily. Building a big data as a service system, that combines big data technologies and cloud computing, will enhance the huge value of big data and tremendously boost the economic growth in various areas. Big data as a service has evolved into a booming market, but with the emergence of larger privacy and security challenges. Privacy and security concerns limit the development of big data as a service and increasingly become one of the main reasons why most data are not shared and well utilized. This dissertation aims to build a new incrementally deployable middleware for the current and future big data as a service eco-system in order to guarantee privacy and security. This middleware will retain privacy and security in the data querying and ensure privacy preservation in data analysis. In addition, emerging cloud computing contributes to providing valuable services associated with machine learning (ML) techniques. We consider privacy issues in both traditional queries and ML queries (i.e., ML classification) in this dissertation. The final goal is to design and develop a demonstrable system that can be deployed in the big data as a service system in order to guarantee the privacy of data/ service owners as well as users, enabling secure data analysis and services.Firstly, we consider a private dataset composed of a set of individuals, and the data is outsourced to a remote cloud server. We revisit the classic query auditing problem in the outsourcing scenario. Secondly, we study privacy preserving neural network classification where source data is randomly partitioned. Thirdly, we concern the privacy of confidential training dataset and models which are typically trained in a centralized cloud server but publicly accessible, \ie online ML-as-a-Service (MLaaS). Lastly, we consider the offline MLaaS systems. We design, implement, and evaluate a secure ML framework to enable MLaaS on clients' edge devices, where a ``encrypted'' ML models are stored locally.
Show less
- Title
- Investigating anti-biofilm and anti-persister activities of natural compounds and antimicrobial proteins
- Creator
- Jin, Xing
- Date
- 2020
- Description
-
Bacterial biofilm formation is frequently involved in the development of chronic infectious diseases. Inhibiting biofilms is challenging due...
Show moreBacterial biofilm formation is frequently involved in the development of chronic infectious diseases. Inhibiting biofilms is challenging due to their tolerance against conventional antibiotics which are not effective to penetrating biofilm matrix to kill the cells residing in biofilms. Metabolically dormant cells known as persisters are also not eradicated by antibiotic treatment. Therefore, novel antimicrobial drugs that can kill non-growing persisters or inhibit biofilms are needed urgently. Here, we investigate the anti-biofilm and anti-persister activities of new drug candidates including plant extracts, fatty acids and colicins. We firstly screened 50 different plant extracts on enterohemorrhagic E. coli and Listeria monocytogenes, and identified Cancavalia ensiformis-derived lectin Concanavalin A (ConA) inhibits biofilm formation of enterohemorrhagic E. coli and Listeria monocytogenes by binding to carbohydrates on bacterial cell surface. Biofilm results support that ConA lectin can be applied for developing anti-adherent and anti-biofilm agents to control biofilms. Also, fatty acids may be promising candidates as anti-persister or anti-biofilm agents, because some fatty acids exhibit antimicrobial effects. We screened a fatty acid library consisting of 65 different fatty acid molecules for altered persister formation. We found that undecanoic acid, lauric acid, and N-tridecanoic acid inhibited E. coli persister cell formation including enterohemorrhagic E. coli EDL933. These fatty acids were all medium chain saturated forms. Furthermore, the fatty acids repressed EHEC biofilm formation (for example, by 8-fold for lauric acid) without having antimicrobial activity. This study demonstrates that medium chain saturated fatty acids can serve as anti-persister and anti-biofilm agents that may be applied to treat bacterial infections. Colicins, a type of antimicrobial bacteriocins, are considered as a viable alternative of conventional antibiotics due to their unique cell killing mechanisms that can damage cells by pore-forming on the cell membrane, nuclease activity, and cell wall synthesis inhibition. In this study, we utilized cell-free protein synthesis to produce colicins with different modes of action. We optimized the production yield and activity of colicins in cell-free system. Also, we tested effect of cell-free produced colicins on persister cell formation and biofilm formation. We illustrated that colicins kill persister cells and biofilm cells. Moreover, colicins produced from the engineered probiotic E. coli cells, which can be used as a living medicine, specifically and significantly eradicate target biofilms without affecting other bacterial population. Colicins have great potential to be an antibiotic alternative, and engineered probiotic E. coli is a potential candidate for engineered bacterial therapeutics.
Show less
- Title
- DIAGNOSING AND TREATING ADHD: CLINICIAN CHARACTERISTICS, METHODS OF DIAGNOSIS, DIAGNOSTIC RATES, AND TREATMENT RECOMMENDATIONS
- Creator
- Haak, Christopher Luke
- Date
- 2019
- Description
-
Attention-deficit/hyperactivity disorder (ADHD) is one of the top five most common referrals among all neuropsychologists (Sweet et al. 2015)...
Show moreAttention-deficit/hyperactivity disorder (ADHD) is one of the top five most common referrals among all neuropsychologists (Sweet et al. 2015) and continues to elicit public and professional concern about over-diagnosis in children (Sciutto & Eisenberg, 2007) and under-diagnosis in adults (Asheron et al., 2012; Kooji et al., 2010). In recent years, the prevalence of ADHD has increased (Polanczyk et al., 2007 & 2014, Thomas et al., 2015). It is unclear what is driving these changes though changes in criteria may be playing a role (van de Voort et al., 2014). Further, there has been little research on whether professional training, beliefs, and practice factors can influence the likelihood to diagnose ADHD. The purpose of this study was to examine the extent to which neuropsychologists’ professional characteristics, training, and beliefs about ADHD diagnosis and treatment influence their likelihood to diagnose ADHD. The study also evaluated whether there are differences in assessing and treating ADHD based upon the client population focus (child, lifespan, or adult) of neuropsychologists. Participants in this study were 106 neuropsychologists from across the United States and Canada who were recruited through neuropsychology listservs to participate in an online survey. Results indicated that population focus was associated with significant differences in approach to diagnosing and treating ADHD, with child- and lifespan-focused neuropsychologists reporting higher rates of ADHD diagnosis. Additionally, having a higher percent of clinical cases in which ADHD is a referral question and greater self-reported adherence to following full diagnostic criteria for making a diagnosis were associated with higher ADHD diagnostic rates, controlling for age, gender, ethnicity, and other professional characteristics. This study is among the first to examine specific clinician factors impacting diagnostic rates and its findings have several implications for practice and research.
Show less
- Title
- Systematic Serendipity: A Study in Discovering Anomalous Astrophysics
- Creator
- Giles, Daniel K
- Date
- 2020
- Description
-
In the present era of large scale surveys, big data presents new challenges to the discovery process for anomalous data. Advances in astronomy...
Show moreIn the present era of large scale surveys, big data presents new challenges to the discovery process for anomalous data. Advances in astronomy are often driven by serendipitous discoveries. Such data can be indicative of systematic errors, extreme (or rare) forms of known phenomena, or most interestingly, truly novel phenomena which exhibit as-of-yet unobserved behaviors. As survey astronomy continues to grow, the size and complexity of astronomical databases will increase, and the ability of astronomers to manually scour data and make such discoveries decreases. In this work, we introduce a machine learning-based method to identify anomalies in large datasets to facilitate such discoveries, and apply this method to long cadence light curves from NASA's Kepler Mission. Our method clusters data based on density, identifying anomalies as data that lie outside of dense regions in a derived feature space. First we present a proof-of-concept case study and we test our method on four quarters of the Kepler long cadence light curves. We use Kepler's most notorious anomaly, Boyajian's Star (KIC 8462852), as a rare `ground truth' for testing outlier identification to verify that objects of genuine scientific interest are included among the identified anomalies. Additionally, we report the full list of identified anomalies for these quarters, and present a sample subset of identified outliers that includes unusual phenomena, objects that are rare in the Kepler field, and data artifacts. By identifying <4% of each quarter as outlying data, under 6k individual targets for the dataset used, we demonstrate that this anomaly detection method can create a more targeted approach in searching for rare and novel phenomena.We further present an outlier scoring methodology to provide a framework of prioritization of the most potentially interesting anomalies. We have developed a data mining method based on k-Nearest Neighbor distance in feature space to efficiently identify the most anomalous light curves. We test variations of this method including using principal components of the feature space, removing select features, the effect of the choice of k, and scoring to subset samples. We evaluate the performance of our scoring on known object classes and find that our scoring consistently scores rare (<1000) object classes higher than common classes, meaning that rarer objects are successfully prioritized over common objects. The most common class, categorized as miscellaneous stars without any major variability, and rotational variables compose well over two-thirds of the KIC, yet are considerably underrepresented in the top outliers. We have applied scoring to all long cadence light curves of quarters 1 to 17 of Kepler's prime mission and present outlier scores for all 2.8 million light curves for the roughly 200k objects.
Show less
- Title
- Characterization of turbulent mixing near roadways based on measurements of short-term turbulence kinetic energy and traffic conditions
- Creator
- Hu, Zhice
- Date
- 2020
- Description
-
Turbulence determines how vehicle emissions mix with the surrounding air and determine the distribution of pollutants on the roadway and...
Show moreTurbulence determines how vehicle emissions mix with the surrounding air and determine the distribution of pollutants on the roadway and downwind. 600 5-min near roadway simultaneous measurements (2016 to 2018) of turbulent kinetic energy (TKE), meteorological conditions, and traffic information (vehicle flow rate, density, and traffic mix (LDVs & HDVs) were used to characterize TKE. Short-term measurements (5 min.) were required to characterize the large variation in traffic flow rate that occurred in short time periods. Two roadways (Lakeshore Drive (LSD), Dan Ryan Expressway (DRE)) with distinctly different traffic composition (HDV%) and road configurations were selected for monitoring. Results indicate that variations in near-road wind speed (0.5 to 3.5 m s-1) had only a slight influence on TKE measurements. Background contributed 40% of the total measured TKE. The average dissipation rate traffic-induced TKE from on-road to near-road measurement was 90%. The average near roadway TKE (background subtracted) was 0.6 (m2 s-2) (0.2 st. dev) for LDVs only, and 0.8 (m2 s-2) (0.3 st. dev) for mixed fleet traffic flow (HDV averaged 8.4%). The increase in TKE was related to the increase in the HDV flow rate for free-flow traffic conditions but not for congestion conditions. TKE generated by individual HDVs was significantly higher than TKE generated from individual LDVs for free-flow traffic conditions. HDVs represent only a small fraction of the vehicle fleet mix (typical 1 to 10%) so that the overall effect of HDVs in changing vehicle fleet is difficult to quantify. However, the single HDV can induce near 11 times TKE than a single LDV in free-flow condition, which can validate the significant variation in the ensemble mean traffic-induced TKE under the same traffic fleet flow that is due to HDVs.
Show less
- Title
- Defense-in-Depth for Cyber-Secure Network Architectures of Industrial Control Systems
- Creator
- Arnold, David James
- Date
- 2024
- Description
-
Digitization and modernization efforts have yielded greater efficiency, safety, and cost-savings for Industrial Control Systems (ICS). To...
Show moreDigitization and modernization efforts have yielded greater efficiency, safety, and cost-savings for Industrial Control Systems (ICS). To achieve these gains, the Internet of Things (IoT) has become an integral component of network infrastructures. However, integrating embedded devices expands the network footprint and softens cyberattack resilience. Additionally, legacy devices and improper security configurations are weak points for ICS networks. As a result, ICSs are a valuable target for hackers searching for monetary gains or planning to cause destruction and chaos. Furthermore, recent attacks demonstrate a heightened understanding of ICS network configurations within hacking communities. A Defense-in-Depth strategy is the solution to these threats, applying multiple security layers to detect, interrupt, and prevent cyber threats before they cause damage. Our solution detects threats by deploying an Enhanced Data Historian for Detecting Cyberattacks. By introducing Machine Learning (ML), we enhance cyberattack detection by fusing network traffic and sensor data. Two computing models are examined: 1) a distributed computing model and 2) a localized computing model. The distributed computing model is powered by Apache Spark, introducing redundancy for detecting cyberattacks. In contrast, the localized computing model relies on a network traffic visualization methodology for efficiently detecting cyberattacks with a Convolutional Neural Network. These applications are effective in detecting cyberattacks with nearly 100% accuracy. Next, we prevent eavesdropping by applying Homomorphic Encryption for Secure Computing. HE cryptosystems are a unique family of public key algorithms that permit operations on encrypted data without revealing the underlying information. Through the Microsoft SEAL implementation of the CKKS algorithm, we explored the challenges of introducing Homomorphic Encryption to real-world applications. Despite these challenges, we implemented two ML models: 1) a Neural Network and 2) Principal Component Analysis. Finally, we hinder attackers by integrating a Cyberattack Lockdown Network with Secure Ultrasonic Communication. When a cyberattack is detected, communication for safety-critical elements is redirected through an ultrasonic communication channel, establishing physical network segmentation with compromised devices. We present proof-of-concept work in transmitting video via ultrasonic communication over an Aluminum Rectangular Bar. Within industrial environments, existing piping infrastructure presents an optimal solution for cost-effectively preventing eavesdropping. The effectiveness of these solutions is discussed within the scope of the nuclear industry.
Show less