Recent Submissions

  • Two-scale Homogenization and Numerical Methods for Stationary Mean-field Games

    Yang, Xianjin (2020-07) [Dissertation]
    Advisor: Gomes, Diogo A.
    Committee members: Shamma, Jeff S.; Parsani, Matteo; Achdou, Yves
    Mean-field games (MFGs) study the behavior of rational and indistinguishable agents in a large population. Agents seek to minimize their cost based upon statis- tical information on the population’s distribution. In this dissertation, we study the homogenization of a stationary first-order MFG and seek to find a numerical method to solve the homogenized problem. More precisely, we characterize the asymptotic behavior of a first-order stationary MFG with a periodically oscillating potential. Our main tool is the two-scale convergence. Using this convergence, we rigorously derive the two-scale homogenized and the homogenized MFG problems. Moreover, we prove existence and uniqueness of the solution to these limit problems. Next, we notice that the homogenized problem resembles the problem involving effective Hamiltoni- ans and Mather measures, which arise in several problems, including homogenization of Hamilton–Jacobi equations, nonlinear control systems, and Aubry–Mather theory. Thus, we develop algorithms to solve the homogenized problem, the effective Hamil- tonians, and Mather measures. To do that, we construct the Hessian Riemannian flow. We prove the convergence of the Hessian Riemannian flow in the continuous setting. For the discrete case, we give both the existence and the convergence of the Hessian Riemannian flow. In addition, we explore a variant of Newton’s method that greatly improves the performance of the Hessian Riemannian flow. In our numerical experiments, we see that our algorithms preserve the non-negativity of Mather mea- sures and are more stable than related methods in problems that are close to singular. Furthermore, our method also provides a way to approximate stationary MFGs.
  • Design and Synthesis of MXene Derived Materials for Advanced Electronics and Energy Harvesting Applications

    Tu, Shao Bo (2020-06-09) [Dissertation]
    Advisor: Zhang, Xixiang
    Committee members: Alshareef, Husam N.; Ooi, Boon S.; Li, Xiaohang; Xu, Bin
    In this thesis, we capitalize on the two-dimensional (2D) nature of MXenes by using them as precursors for the synthesis of 2D functional material. MXenes are easily intercalated with monovalent cations K, Na, Li due to their expanded d-spacing after etching. Based on these ideas, we have developed new synthesis processes of texture functional materials using MXenes as precursors. We have successfully synthesized two-dimensional Nb2C MXene based high aspect ratio ferroelectric potassium niobate (KNbO3) and well-oriented photoluminescent rare earth doped lithium niobate (LiNbO3:Pr3+) crystals, which have great potential in opto-electronics applications. In addition, this thesis demonstrates that poly(vinylidene fluoride) (PVDF)-based percolative composites using two-dimensional (2D) MXene nanosheets as fillers exhibit significantly enhanced dielectric permittivity. Furthermore, we fabricated MXene/in-plane aligned PVDF photo-thermo-mechanical solar tracking actuator for energy harvesting applications.
  • Bacterial Endophytes from Pioneer Desert Plants for Sustainable Agriculture

    Eida, Abdul Aziz (2020-06) [Dissertation]
    Advisor: Hirt, Heribert
    Committee members: Saad, Maged M.; Pain, Arnab; Aranda, Manuel; Kopriva, Stanislav
    One of the major challenges for agricultural research in the 21st century is to increase crop productivity to meet the growing demand for food and feed. Biotic (e.g. plant pathogens) and abiotic stresses (e.g. soil salinity) have detrimental effects on agricultural productivity, with yield losses being as high as 60% for major crops such as barley, corn, potatoes, sorghum, soybean and wheat, especially in semi-arid regions such as Saudi Arabia. Plant growth promoting bacteria isolated from pioneer desert plants could serve as an eco-friendly, sustainable solution for improving plant growth, stress tolerance and health. In this dissertation, culture-independent amplicon sequencing of bacterial communities revealed how native desert plants influence their surrounding bacterial communities in a phylogeny-dependent manner. By culture-dependent isolation of the plant endosphere compartments and a number of bioassays, more than a hundred bacterial isolates with various biochemical properties, such as nutrient acquisition, hormone production and growth under stress conditions were obtained. From this collection, five phylogenetically diverse bacterial strains were able to promote the growth of the model plant Arabidopsis thaliana under salinity stress conditions in a common mechanism of inducing transcriptional changes of tissue-specific ion transporters and lowering Na+/K+ ratios in the shoots. By combining a number of in vitro bioassays, plant phenotyping and volatile-mediated inhibition assays with next-generation sequencing technology, gas chromatography–mass spectrometry and bioinformatics tools, a candidate strain was presented as a multi-stress tolerance promoting bacterium with potential use in agriculture. Since recent research showed the importance of microbial partners for enhancing the growth and health of plants, a review of the different factors influencing plant-associated microbial communities is presented and a framework for the successful application of microbial inoculants in agriculture is proposed. The presented work demonstrates a holistic approach for tackling agricultural challenges using microbial inoculants from desert plants by combining culturomics, phenomics, genomics and transcriptomics. Microbial inoculants are promising tools for studying abiotic stress tolerance mechanisms in plants, and they provide an eco-friendly solution for increasing crop yield in arid and semi-arid regions, especially in light of a dramatically growing human population and detrimental effects of global warming and climate change.
  • Hybrid Local/Nonlocal Continuum Mechanics Modeling and Simulation for Material Failure

    Wang, Yongwei (2020-06) [Dissertation]
    Advisor: Lubineau, Gilles
    Committee members: Thoroddsen, Sigurdur T.; Hoteit, Ibrahim; Florentin, Eric
    The classical continuum mechanics, which studies the mechanical behavior of structures based on partial differential equations, shows its deficiencies when it encounters a discontinuity. Peridynamics based on integral equations can simulate fracture but suffers from high computational costs. A hybrid local/nonlocal model combining the advantages of peridynamics with those of classical continuum mechanics can simulate fracture and reduce the computational cost. Under the framework of the hybrid local/nonlocal model, this research developed an approach and computational codes for fracture simulations. First, we developed the computational codes based on the hybrid model with a priori partition of the domain between local and nonlocal models to tackle engineering problems with relevant level of difficulty. Second, we developed a strength-induced approach to enhance the capability of the computational codes because the strength-induced approach can limit the peridynamic model to necessary computational steps at the time level and a relatively small zone at the space level during a simulation. The strength-induced approach also improved the hybrid models by enabling an automatic partition of the domain without manual involvement. At last, a strength-induced computational code was developed based on this research. This dissertation complemented and illustrated numerically some previous work of Cohmas laboratory, in which a new route was introduced toward simulating the whole process of material behaviors including elastic deformation, crack nucleation and propagation until structural failure.
  • Whole Genome Sequencing as a Tool to Study the Genomic Landscape of Pathogens

    Hala, Sharif (2020-06) [Dissertation]
    Advisor: Pain, Arnab
    Committee members: Merzaban, Jasmeen; Khashab, Niveen; Carr , Michael
    In healthcare settings and beyond, the ESKAPE pathogens (Enterococcus faecium, Staphylococcus aureus, Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa, and Enterobacter species) among other pathogens exchange antibiotic resistance and virulence factors and emerge as new infectious clones. According to the Saudi General Authority for Statistics (stats.gov.sa), Saudi Arabia is a country where more than 27 million pilgrims meet in annual continual mass-gathering events. This massive influx of people could introduce novel pathogens to the community that could not necessarily be detected with traditional culture-dependent clinical microbiological tests. Conventional clinical microbiology and environmental pathogen detection methods have had many limitations and narrow search scope. These methods can only target known and culturable pathogens. Over the past decade, applications of next-generation sequencing (NGS) and bioinformatics tools have revolutionized the way pathogens are detected and their relevant phenotypes such as clonal types, antibiotic resistance are predicted to aid in clinical decision making as additional practice to traditional clinical microbiology-based testing protocols. The aim of this study was to apply whole-genome sequencing (WGS) and bioinformatic analysis tools on clinical samples and bacterial isolates in order to pave the way for transforming current clinical microbiology practices in a tertiary referral hospital in Jeddah, Saudi Arabia. My attempt to utilize WGS as a tool on pathogenic strains in this study combined with the clinical data has resulted in discovering a silent outbreak of an emerging hypervirulent strain of Klebsiella pneumoniae (Chapter 2). Analysis of the strains antimicrobial profiles genetically has yielded the first characterization of a misidentified Klebsiella quasipneumoniae harboring plasmid-mediated carbapenemases of Klebsiella pneumoniae carbapenemases (KPC) (Chapter 3). Similarly, I was able to study mobile colistin resistance genes in the isolates and identify a novel occurrence of mcr-1 and mcr-8 (Chapter 4). I applied clinical metagenomic protocol on an intestinal biopsy of an inflammatory bowel disease patient with Crohn’s disease, where I identified an association of three co-occurring and an actively replicating non-tuberculosis mycobacteria (Chapter 5). The deployment of whole-genome sequencing and metagenomic in infectious disease surveillance and diagnostics could prove beneficial in limiting epidemics and detect transmission patterns of antimicrobial-resistant genes.
  • Ultraviolet Band Based Underwater Wireless Optical Communication

    Sun, Xiaobin (2020-05) [Dissertation]
    Advisor: Ooi, Boon S.
    Committee members: Shamma, Jeff S.; Jones, Burton; Peng, Gang Ding
    Underwater wireless optical communication (UWOC) has attracted increasing interest for data transfer in various underwater activities. However, the complexity of the water environment poses considerable challenges to establish aligned and reliable UWOC links. Therefore, solutions that are capable of relieving the requirements on positioning, acquisition and tracking (PAT) are highly demanded. Different from the conventional blue-green light band utilized in UWOC, ultraviolet (UV) light is featured with low solar background noise, non-line-of-sight (NLOS) and good secrecy. The proposed work is directed towards the demonstration and evaluating the feasibility of high- speed NLOS UWOC for easing the strict requirement on alignment, and thus circumvent the issues of scintillation, deep-fading, and complete signal blockage presented in conventional LOS UWOC. This work was first started with the investigation of proper NLOS configurations. Path loss (PL) was chosen as a figure-of-merit for link performance. With the understanding of favorable NLOS UWOC configurations, we established a 377-nm laser-based, the first-of-its-kind NLOS UWOC link. The practicality of such NLOS UWOC links has been further tested in a field trial. Besides the underwater communication links, UV-based NLOS is also appealing to be the link for direct communication across the wavy water-air interface. Investigations for such a direct communication link have been carried out to study data rate, coverage and robustness to the dynamic wave movement, based on the performance of different modulation schemes, including non-return-to-zero (NRZ)-OOK and quadrature amplitude modulation (QAM)-orthogonal frequency division multiplexing (OFDM). Further this study, an in-Red Sea canal field in-situ test has been conducted, showing strong robustness of the system. In addition, an in-diving pool drone-aided real-application deployment has been carried on. The trial results indicate link stability, which alleviates the issues brought about by the misalignment and mobility in harsh environments, paving the way towards real applications. Our studies pave the way foreventual applications of UWOC byrelieving the strict requirements on PAT using UV-based NLOS. Such modality is much sought-after for implementing robust, secure, and high-speed UWOC links in harsh oceanic environments.
  • A Green and Powerful Method toward Well-defined Polycarbonates and Polycarbonate-Based Block Copolymers from CO2 and Epoxides

    Alzahrany, Yahya (2020-05) [Dissertation]
    Advisor: Hadjichristidis, Nikos
    Committee members: Nunes, Suzana Pereira; Bakr, Osman; Avgeropoulos, Apostolos
    The use of waste gas such as carbon dioxide (CO2) to prepare useful and valuable polymers benefits both the economy and the environment. Various strategies have been developed to reduce CO2 emission as well as to transfer CO2 into high-value products. CO2/epoxide copolymerization is one of the most promising methods of not only reducing the CO2 emission from the atmosphere but also producing biodegradable CO2-based materials that are CO2 as source-abundant, renewable, cheap, non-flammable and non-toxic. However, the activation of CO2 is one of several problems associated with the polymerization of CO2 due to its stability as a thermodynamic end product. Herein, my dissertation describes the effectiveness of new lithium/phosphazene complexes to generate highly active species for CO2/epoxide copolymerization and to capture/activate CO2 molecules for the nucleophilic attack of the active species. Well-defined polycarbonates and polycarbonate-based block copolymers are produced that have control of molecular weights, unimodal distributions and narrow molecular weight distributions (Chapter 3 and 4). Besides, these complexes provide access to prepare CO2-based triblock copolymers that are powerful candidates to serve as the next generation of thermoplastic elastomers (Chapter 4). Additionally, these complexes are applied for the anionic polymerization of petrochemical-based sources such as styrene and dienes producing polymers in faster rate of polymerization with control of molecular characteristics (Chapter 2). A general introduction of polymers and their classification based on composition, chemical structure, mechanical properties, degradability, source, applications, and preparative methods, is covered in Chapter 1
  • Advanced Sediment Characterization

    Salva Ramirez, Marisol (2020-05) [Dissertation]
    Advisor: Santamarina, Juan Carlos
    Committee members: Ki Kim, Hyun; Vahrenkamp., Volker; Ahmed, Shehab
    Soil data accumulated during the last century and more recent developments in sensors and information technology prompt the development of new geotechnical solutions for soil assessment. We have advanced three complementary tools: Lab-on-a-Bench, the soil properties database with corresponding IT Tool and the in-situ characterization Multiphysics Probe. Lab-on-a-Bench technology combines cutting-edge sensors and sensing concepts within compact devices and effective laboratory protocols to allow multi physics soil characterization: specific surface measurements for fine-grained soils and particle size distribution, shape, packing densities and angle of repose for coarse-grained soils using image analysis and corresponding devices. The soil properties database and complementary IT Tool provide a self-consistent set of soil parameters based on known properties. The advances in in-situ characterization focus on a Multiphysics Probe and include measurements of remnant magnetization to identify metalliferous sediments for deep-sea mining applications and shear wave measurements for stiffness assessments. All methods, protocols, devices and technology are applied to Red Sea sediments to establish a baseline for future industrial and economic developments
  • Hierarchical Matrix Operations on GPUs

    Boukaram, Wagih Halim (2020-04-26) [Dissertation]
    Advisor: Keyes, David E.
    Committee members: Ketcheson, David I.; Hadwiger, Markus; Turkiyyah, George; Darve, Eric F.
    Large dense matrices are ubiquitous in scientific computing, arising from the discretization of integral operators associated with elliptic pdes, Schur complement methods, covariances in spatial statistics, kernel-based machine learning, and numerical optimization problems. Hierarchical matrices are an efficient way for storing the dense matrices of very large dimension that appear in these and related settings. They exploit the fact that the underlying matrices, while formally dense, are data sparse. They have a structure consisting of blocks many of which can be well-approximated by low rank factorizations. A hierarchical organization of the blocks avoids superlinear growth in memory requirements to store n × n dense matrices in a scalable manner, requiring O(n) units of storage with a constant depending on a representative rank k for the low rank blocks. The asymptotically optimal storage requirement of the resulting hierarchical matrices is a critical advantage, particularly in extreme computing environments, characterized by low memory per processing core. The challenge then becomes to develop the parallel linear algebra operations that can be performed directly on this compressed representation. In this dissertation, I implement a set of hierarchical basic linear algebra subroutines (HBLAS) optimized for GPUs, including hierarchical matrix vector multiplication, orthogonalization, compression, low rank updates, and matrix multiplication. I develop a library of open source batched kernel operations previously missing on GPUs for the high performance implementation of the H2 operations, while relying wherever possible on existing open source and vendor kernels to ride future improvements in the technology. Fast marshaling routines extract the batch operation data from an efficient representation of the trees that compose the hierarchical matrices. The methods developed for GPUs extend to CPUs using the same code base with simple abstractions around the batched routine execution. To demonstrate the scalability of the hierarchical operations I implement a distributed memory multi-GPU hierarchical matrix vector product that focuses on reducing communication volume and hiding communication overhead and areas of low GPU utilization using low priority streams. Two demonstrations involving Hessians of inverse problems governed by pdes and space-fractional diffusion equations show the effectiveness of the hierarchical operations in realistic applications.
  • Improved Design of Quadratic Discriminant Analysis Classi er in Unbalanced Settings

    Bejaoui, Amine (2020-04-23) [Dissertation]
    Advisor: Alouini, Mohamed Slim
    Committee members: Huser, Raphaël G.; Kammoun, Abla
    The use of quadratic discriminant analysis (QDA) or its regularized version (RQDA) for classi cation is often not recommended, due to its well-acknowledged high sensitivity to the estimation noise of the covariance matrix. This becomes all the more the case in unbalanced data settings for which it has been found that R-QDA becomes equivalent to the classi er that assigns all observations to the same class. In this paper, we propose an improved R-QDA that is based on the use of two regularization parameters and a modi ed bias, properly chosen to avoid inappropriate behaviors of R-QDA in unbalanced settings and to ensure the best possible classi cation performance. The design of the proposed classi er builds on a re ned asymptotic analysis of its performance when the number of samples and that of features grow large simultaneously, which allows to cope e ciently with the high-dimensionality frequently met within the big data paradigm. The performance of the proposed classi er is assessed on both real and synthetic data sets and was shown to be much higher than what one would expect from a traditional R-QDA.
  • Computation of High-Dimensional Multivariate Normal and Student-t Probabilities Based on Matrix Compression Schemes

    Cao, Jian (2020-04-22) [Dissertation]
    Advisor: Genton, Marc G.
    Committee members: Keyes, David E.; Rue, Haavard; Panaretos, Victor
    The first half of the thesis focuses on the computation of high-dimensional multivariate normal (MVN) and multivariate Student-t (MVT) probabilities. Chapter 2 generalizes the bivariate conditioning method to a d-dimensional conditioning method and combines it with a hierarchical representation of the n × n covariance matrix. The resulting two-level hierarchical-block conditioning method requires Monte Carlo simulations to be performed only in d dimensions, with d ≪ n, and allows the dominant complexity term of the algorithm to be O(n log n). Chapter 3 improves the block reordering scheme from Chapter 2 and integrates it into the Quasi-Monte Carlo simulation under the tile-low-rank representation of the covariance matrix. Simulations up to dimension 65,536 suggest that this method can improve the run time by one order of magnitude compared with the hierarchical Monte Carlo method. The second half of the thesis discusses a novel matrix compression scheme with Kronecker products, an R package that implements the methods described in Chapter 3, and an application study with the probit Gaussian random field. Chapter 4 studies the potential of using the sum of Kronecker products (SKP) as a compressed covariance matrix representation. Experiments show that this new SKP representation can save the memory footprint by one order of magnitude compared with the hierarchical representation for covariance matrices from large grids and the Cholesky factorization in one million dimensions can be achieved within 600 seconds. In Chapter 5, an R package is introduced that implements the methods in Chapter 3 and show how the package improves the accuracy of the computed excursion sets. Chapter 6 derives the posterior properties of the probit Gaussian random field, based on which model selection and posterior prediction are performed. With the tlrmvnmvt package, the computation becomes feasible in tens of thousands of dimensions, where the prediction errors are significantly reduced.
  • Predicting Gene Functions and Phenotypes by combining Deep Learning and Ontologies

    Kulmanov, Maxat (2020-04-08) [Dissertation]
    Advisor: Hoehndorf, Robert
    Committee members: Arold, Stefan T.; Moshkov, Mikhail; Hunter, Larry
    The amount of available protein sequences is rapidly increasing, mainly as a consequence of the development and application of high throughput sequencing technologies in the life sciences. It is a key question in the life sciences to identify the functions of proteins, and furthermore to identify the phenotypes that may be associated with a loss (or gain) of function in these proteins. Protein functions are generally determined experimentally, and it is clear that experimental determination of protein functions will not scale to the current { and rapidly increasing { amount of available protein sequences (over 300 million). Furthermore, identifying phenotypes resulting from loss of function is even more challenging as the phenotype is modi ed by whole organism interactions and environmental variables. It is clear that accurate computational prediction of protein functions and loss of function phenotypes would be of signi cant value both to academic research and to the biotechnology industry. We developed and expanded novel methods for representation learning, predicting protein functions and their loss of function phenotypes. We use deep neural network algorithm and combine them with symbolic inference into neural-symbolic algorithms. Our work signi cantly improves previously developed methods for predicting protein functions through methodological advances in machine learning, incorporation of broader data types that may be predictive of functions, and improved systems for neural-symbolic integration. The methods we developed are generic and can be applied to other domains in which similar types of structured and unstructured information exist. In future, our methods can be applied to prediction of protein function for metagenomic samples in order to evaluate the potential for discovery of novel proteins of industrial value. Also our methods can be applied to the prediction of loss of function phenotypes in human genetics and incorporate the results in a variant prioritization tool that can be applied to diagnose patients with Mendelian disorders.
  • Molecular response of a coral reef fish (Acanthochromis polyacanthus) to climate change

    Monroe, Alison (2020-04) [Dissertation]
    Advisor: Voolstra, Christian R.
    Committee members: Magistretti, Pierre; Berumen, Michael L.; Eirin-Lopez, Jose M.; Schunter, Celia
    Marine ecosystems are already threatened by the effects of climate change through increases in ocean temperatures and pCO2 levels due to increasing atmospheric CO2. Marine fish living close to their thermal maximum have been shown to be especially vulnerable to temperatures exceeding that threshold, and even relatively small increases in elevated pCO2 levels have led to behavioral impairments with amplified predation risks. These ongoing threats highlight the need for further understanding of how these changes will impact fish and if any potential for adaptation or acclimation exists. The coral reef fish, Acanthochromis polyacanthus, has been well studied in response to singular environmental changes both through its phenotype and molecular expression profiles within and across generations. However, key questions regarding transgenerational heritability and molecular responses to multiple environmental changes have not been addressed. To further understand A. polyacanthus I examined the mechanisms behind heritability of behavioral tolerance to elevated pCO2 in an attempt to determine the maternal and paternal contributions to this phenotype. There was a strong impact of parental phenotype on the expression profiles of their offspring regardless of environmental exposure. Offspring from both parental pairs expressed mechanisms involved in tolerance to ocean acidification suggesting this phenotype is reliant on input from both parents. Creation of a new proteomic resource, a SWATH spectral library, delivered a closer examination of the link between phenotypic and expression changes. Analysis on different constructed libraries led to the use of an organism whole library combined with study specific data to analyze proteomic changes in A. polyacanthus under the combined environmental changes of ocean acidification and warming. With direct comparisons to transcriptomic changes in the same individuals I identified an additive effect of elevated pCO2 and temperature associated with decreases in growth and development. However, a strong role of parental identity on the expression profiles of offspring reinforced the high genetic variability of this species. This thesis provides novel insights into the heritability of phenotypic traits and the molecular responses to combined stressors in A. polyacanthus, as well as presenting a new resource for proteomic studies in this fish and other non-model species.
  • Efficient and Accurate Numerical Techniques for Sparse Electromagnetic Imaging

    Sandhu, Ali Imran (2020-04) [Dissertation]
    Advisor: Bagci, Hakan
    Committee members: Ooi, Boon S.; Hoteit, Ibrahim; Dorn, Oliver
    Electromagnetic (EM) imaging schemes are inherently non-linear and ill-posed. Albeit there exist remedies to these fundamental problems, more efficient solutions are still being sought. To this end, in this thesis, the non-linearity is tackled in- corporating a multitude of techniques (ranging from Born approximation (linear), inexact Newton (linearized) to complete nonlinear iterative Landweber schemes) that can account for weak to strong scattering problems. The ill-posedness of the EM inverse scattering problem is circumvented by formulating the above methods into a minimization problem with a sparsity constraint. More specifically, four novel in- verse scattering schemes are formulated and implemented. (i) A greedy algorithm is used together with a simple artificial neural network (ANN) for efficient and accu- rate EM imaging of weak scatterers. The ANN is used to predict the sparsity level of the investigation domain which is then used as the L0 - constraint parameter for the greedy algorithm. (ii) An inexact Newton scheme that enforces the sparsity con- straint on the derivative of the unknown material properties (not necessarily sparse) is proposed. The inverse scattering problem is formulated as a nonlinear function of the derivative of the material properties. This approach results in significant spar- sification where any sparsity regularization method could be efficiently applied. (iii) A sparsity regularized nonlinear contrast source (CS) framework is developed to di- rectly solve the nonlinear minimization problem using Landweber iterations where the convergence is accelerated using a self-adaptive projected accelerated steepest descent algorithm. (iv) A 2.5D finite difference frequency domain (FDFD) based in- verse scattering scheme is developed for imaging scatterers embedded in lossy and inhomogeneous media. The FDFD based inversion algorithm does not require the Green’s function of the background medium and appears a promising technique for biomedical and subsurface imaging with a reasonable computational time. Numerical experiments, which are carried out using synthetically generated mea- surements, show that the images recovered by these sparsity-regularized methods are sharper and more accurate than those produced by existing methods. The methods developed in this work have potential application areas ranging from oil/gas reservoir engineering to biological imaging where sparse domains naturally exist.
  • Understanding a Block of Layers in Deep Neural Networks: Optimization, Probabilistic and Tropical Geometric Perspectives

    Bibi, Adel (2020-04) [Dissertation]
    Advisor: Ghanem, Bernard
    Committee members: Heidrich, Wolfgang; Richtarik, Peter; Ma, Yi
    This dissertation aims at theoretically studying a block of layers that is common in al- most all deep learning models. The block of layers of interest is the composition of an affine layer followed by a nonlinear activation that is followed by another affine layer. We study this block from three perspectives. (i) An Optimization Perspective. Is it possible that the output of the forward pass through this block is an optimal solution to a certain convex optimization problem? We show an equivalency between the forward pass through this block and a single iteration of deterministic and stochastic algorithms solving a ten- sor formulated convex optimization problem. As consequence, we derive for the first time a formula for computing the singular values of convolutional layers surpassing the need for the prohibitive construction of the underlying linear operator. Thereafter, we show that several deep networks can have this block replaced with the corresponding optimiza- tion algorithm predicted by our theory resulting in networks with improved generalization performance. (ii) A Probabilistic Perspective. Is it possible to analytically analyze the output of a deep network upon subjecting the input to Gaussian noise? To that regard, we derive analytical formulas for the first and second moments of this block under Gaussian input noise. We demonstrate that the derived expressions can be used to efficiently analyze the output of an arbitrary deep network in addition to constructing Gaussian adversarial attacks surpassing any need for prohibitive data augmentation procedures. (iii) A Tropi- cal Geometry Perspective. Is it possible to characterize the decision boundaries of this block as a geometric structure representing a solution set to a certain class of polynomials (tropical polynomials)? If so, then, is it possible to utilize this geometric representation of the decision boundaries for novel reformulations to classical computer vision and machine learning tasks on arbitrary deep networks? We show that the decision boundaries of this block are a subset of a tropical hypersurface, which is intimately related to a the polytope that is the convex hull of two zonotopes. We utilize this geometric characterization to shed lights on new perspectives of network pruning.
  • Fabrication and Characterization of GaN-Based Superluminescent Diode for Solid-State Lighting and Visible Light Communication

    Alatawi, Abdullah (2020-04) [Dissertation]
    Advisor: Ooi, Boon S.
    Committee members: Ohkawa, Kazuhiro; ABDELSABOOR, Omar Mohammed; Zhao, HongPing
    To date, group-III-nitride has undergone continuous improvements to provide a broader range of industrial applications, such as solid-state lighting (SSL), visible light communications (VLC), and light projection. Recently, VLC has attained substantial attention in the field of wireless communication because it offers ~ 370 THz of bandwidth of unregulated visible spectrum, which makes it a critical factor in the evolution of the 5G networks and beyond. GaN-based light-emitting diode (LED) and laser diode (LD) have become increasingly appealing in energy-sufficient SSL replacing conventional light sources. However, III- nitride LEDs suffer from efficiency-droop in their external quantum efficiency associated with high current densities, and their modulation bandwidth is limited to 10 ~ 100 MHz. Although LDs have shown gigabit-modulation bandwidth, unfavorable artifacts, such as speckles are observed, which may raise a concern about eye safety. This dissertation is devoted to the fabrication and electrical and optical characterization of a new class of III-nitride light-emitter known as superluminescent diode (SLD). SLD works in an amplified spontaneous emission (ASE) regime, and it combines several advantages from both LD and LED, such as droop-free, speckle-free, low-spatial coherence, broader emission, high-optical power, and directional beam. Here, SLDs were fabricated by a focused ion beam by tilting the front facet of the waveguide to suppress the lasing mode. They showed a high-power of 474 mW on c-plane GaN-substrate with a large spectral bandwidth of 6.5 nm at an optical power of 105 mW. To generate SLD- based white light, a YAG-phosphor-plate was integrated, and a CRI of 85.1 and CCT of 3392 K were measured. For the VLC link, SLD showed record high-data rates of 1.45 Gbps and 3.4 Gbps by OOK and DMT modulation schemes, respectively. Additionally, a widely single- and dual-wavelength tunability were designed using SLD-based external cavity (SLD-EC) configuration for a tunable blue laser source. These results underscore the practicality of c-plane SLDs in realizing high-power, high data rate, speckle-free, and droop-free SSL-VLC apparatus. Additionally, the SLD-EC configuration allows a wide range of applications, including biomedical applications, optical communication, and high-resolution spectroscopy.
  • Theoretical and experimental investigation of liquid droplets flashing for low cost seawater desalination

    Alrowais, Raid (2020-04) [Dissertation]
    Advisor: NG, Kim Choon
    Committee members: Ghaffour, Noreddine; Thoroddsen, Sigurdur T.; Chakraborty, Anutosh
    The high specific energy consumption from all existing seawater desalination methods has heightened the motivation for having more efficient and greener desalination processes to meet the future goals of sustainable seawater desalination. One of the promising thermally-driven desalination methods is the direct-contact spray evaporation and condensation (DCSEC) where the excess enthalpy between feed and equilibrium states of evaporator chambers is exploited with reasonably high flashing efficiency. Further improvements in energy efficacy of DCSEC are boosted by firstly the incorporation of micro/nano-bubbles (M/NB) where micro or nano size subcooled vapor are embedded in the sprayed liquid droplets of evaporator, thereby lowering the temperature brine in evaporator and minimizing the thermal equilibrium effect of brine. The presence of subcooled bubbles increased the available surface area for heat transfer. Secondly, the concept of an evaporator-condenser pair of DCSEC could be extended to a multi-stage arrangement where the latent heat of vapor condensing on the water droplets sprayed within the condenser is recovered. From the experiments, the effect of incorporating the (M/NB) in the DCSEC at optimum feed flow rate yields more than 34% increase in distillate production at feed temperatures greater 47oC and the cooling inlet temperature set at 35oC. The other salient improvement found from the experiments is the increase in performance ratio (PR) up to 3.3 for a 6-stage configuration. This quantum jump in the PR is attributed to the heat recovery effect by as much as 70% of the total heat input. Arising from the DCSEC design, the implicit benefits are the low capital and operational cost, i.e., low CAPEX and OPEX. The former savings is attributed zero physical interfaces such as tube-based heat exchangers or membranes, whilst the latter savings is contributed by significant lesser use of chemicals in the pre-treatment of seawater feed. Lastly, the accompanied benefit is the robustness of the DCSEC processes where it could within stand high salinity of the brine, typically as high as 200,000 ppm.
  • High-Performance Optoelectronics Based on Mixed-Dimensional Organolead Halide Perovskites

    Ma, Chun (2020-04-01) [Dissertation]
    Advisor: Anthopoulos, Thomas D.
    Committee members: Tung, Vincent; McCulloch, Iain; Heeney, Martin
    Halide perovskites have some unique advantages as optoelectronic materials. Metal halide perovskites have been attracting enormous attention for applications in optoelectronic devices such as photodetectors, light-emitting devices and field-effect transistors. The remarkable semiconducting properties have been intensively investigated in recent years. However, the performance of optoelectronics devices based on the conventional perovskite is limited by the ion migration, the mobility of the carriers and the light absorption in the near infrared region and so on. In a decade, numerous attempts are studied to further breakthrough the limitations using both physical and chemical methods. This dissertation is devoted to overcoming the drawbacks by integrating the state-of-art perovskite with other functional materials and to further deciphering the carrier transport mechanics behind the mixed dimensional heterostructures. Field-effect transistors are the workhorse of modern microelectronics. Proof-of-concept devices have been made, utilizing solution-processed perovskite as transistors. Beyond the Field-effect transistors, photodetectors can be construct with a transistor configuration. In this dissertation, we exploited Au dimers with structural darkness to enhance the light harvesting, and utilize sorted semiconducting single-walled carbon nanotubes to enhance the conductivity of thin-film. At last, we developed a hybrid memtransistor, modulable by multiple physical inputs using hybrid perovskite and conjugated polymer heterojunction channels to realize neuromorphic computing.
  • An Experimental Investigation on the Dynamics of Lean Premixed Swirl Flames

    Di Sabatino, Francesco (2020-04) [Dissertation]
    Advisor: Lacoste, Deanna
    Committee members: Roberts, William Lafayette; Knio, Omar; worth, nicolas
    Gas turbine engines are an efficient and flexible way of power generation and aircraft propulsion. Even though different combustion systems can be implemented in these engines, more stringent regulations on pollutant emissions have been imposed throughout the years, especially in regard to nitrogen oxides (NOx). A very promising technology to reduce NOx emissions is lean premixed combustion (LPC), however, it is plagued by intense flame dynamics. Thermoacoustic instabilities, lean blow-off and lean instabilities are examples of dynamical phenomena that are detrimental to the gas turbines. In view of this, the present thesis presents the experimental investigation of the response of lean premixed swirl flames to acoustic perturbations at atmospheric and elevated pressures. The results of this investigation may be used to understand the thermoacoustic instabilities and further could be helpful in their prediction. Moreover, this work addresses the effects of non-thermal plasma discharges on the lean blow-off and stability limits of premixed swirl flames at elevated pressures. For the analysis of the flame response to acoustic fluctuations, the flame transfer functions, the flame dynamics, phase-locked velocity fields, and phase-locked measurements of flame curvature are collected through heat release and velocity fluctuations measurements, phase-locked images of the flame, particle image velocimetry, and planar laser-induced fluorescence, respectively. For the analysis of the effects of plasma discharges on the stability limits, electrical measurements and direct imaging of the flame are performed. The results include the development of an empirical relation based on the laminar burning velocity and on the circulation of the acoustically generated vortex to predict the response of the flame to acoustic fluctuations in different operating conditions. Moreover, the results show that the pressure has a strong impact on the response of lean premixed swirl flames to acoustic oscillations and on the flame-plasma interactions. Therefore, extrapolating results obtained at atmospheric conditions to elevated pressures may result in erroneous conclusions. Furthermore, it is shown that non-thermal plasma discharges can effectively extend the stability limits of lean premixed swirl flames at elevated pressures, underlining the potential of these discharges at conditions relevant for gas turbines.
  • Novel computational methods for promoter identification and analysis

    Umarov, Ramzan (2020-03-02) [Dissertation]
    Advisor: Gao, Xin
    Committee members: Moshkov, Mikhail; Hoehndorf , Robert; Daub, Carsten
    Promoters are key regions that are involved in differential transcription regulation of protein-coding and RNA genes. The gene-specific architecture of promoter sequences makes it extremely difficult to devise a general strategy for their computational identification. Accurate prediction of promoters is fundamental for interpreting gene expression patterns, and for constructing and understanding genetic regulatory networks. In the last decade, genomes of many organisms have been sequenced and their gene content was mostly identified. Promoters and transcriptional start sites (TSS), however, are still left largely undetermined and efficient software able to accurately predict promoters in newly sequenced genomes is not yet available in the public domain. While there are many attempts to develop computational promoter identification methods, reliable tools to analyze long genomic sequences are still lacking. In this dissertation, I present the methods I have developed for prediction of promoters for different organisms. The first two methods, TSSPlant and PromCNN, achieved state-of-the-art performance for discriminating promoter and non-promoter sequences for plant and eukaryotic promoters respectively. For TSSPlant, a large number of features were crafted and evaluated to train an optimal classifier. Prom- CNN was built using a deep learning approach that extracts features from the data automatically. The trained model demonstrated the ability of a deep learning approach to grasp complex promoter sequence characteristics. For the latest method, DeeReCT-PromID, I focus on prediction of the exact positions of the TSSs inside the eukaryotic genomic sequences, testing every possible location. This is a more difficult task, requiring not only an accurate classifier, but also appropriate selection of unique predictions among multiple overlapping high scoring genomic segments. The new method significantly outperform the previous promoter prediction programs by considerably reducing the number of false positive predictions. Specifically, to reduce the false positive rate, the models are adaptively and iteratively trained by changing the distribution of samples in the training set based on the false positive errors made in the previous iteration. The new methods are used to gain insights into the design principles of the core promoters. Using model analysis, I have identified the most important core promoter elements and their effect on the promoter activity. Furthermore, the importance of each position inside the core promoter was analyzed and validated using a large single nucleotide polymorphisms data set. I have developed a novel general approach to detect long range interactions in the input of a deep learning model, which was used to find related positions inside the promoter region. The final model was applied to the genomes of different species without a significant drop in the performance, demonstrating a high generality of the developed method.

View more