Browsing Theses by Issue Date
Now showing items 1-20 of 653
On the Statistical Modeling of the Underwater Optical Wireless Channel Subject to Air Bubbles(2019-05-08) [Thesis]
Advisor: Alouini, Mohamed-Slim
Committee members: Parsani, Matteo; Park, Ki-HongIn underwater wireless optical communications (UWOC), the obstruction of light propagation by air bubbles is one of the main factors which causes light power to fluctuate at the receiver. In this thesis, we construct a statistical model for the received power in the presence of air bubbles. First, we postulate some random variables based on some real experiments, such as the size of a bubble, the generation of each bubble, and the horizontal and vertical movements of a bubble. Second, we mathematically express the amount of obstructed power which the shade of each bubble causes over the beam area and sum them all up to get the total obstructed power. In order to use the method of moments, we find the expectation, the second and/or the third moments of the total obstructed power. Lastly, we use these two or three moments of it to find suitable distributions that match the simulation data, which are the Weibull distribution and Generalized gamma distribution respectively. With these distributions, we construct the statistical model of the received power. Furthermore, we show that those distributions fit well to the simulation data.
A Full Multigrid-Multilevel Quasi-Monte Carlo Approach for Elliptic PDE with Random Coefficients(2019-05-05) [Thesis]
Advisor: Sun, Shuyu
Committee members: Hoteit, Ibrahim; Tempone, Raul; Liu, HailiangThe subsurface flow is usually subject to uncertain porous media structures. However, in most cases we only have partial knowledge about the porous media properties. A common approach is to model the uncertain parameters as random fields, then the expectation of Quantity of Interest(QoI) can be evaluated by the Monte Carlo method. In this study, we develop a full multigrid-multilevel Monte Carlo (FMG-MLMC) method to speed up the evaluation of random parameters effects on single-phase porous flows. In general, MLMC method applies a series of discretization with increasing resolution and computes the QoI on each of them, the success of which lies in the effective variance reduction. We exploit the similar hierarchies of MLMC and multigrid methods, and obtain the solution on coarse mesh Qcl as a byproduct of the multigrid solution on fine mesh Qfl on each level l. In the cases considered in this thesis, the computational saving is 20% theoretically. In addition, a comparison of Monte Carlo and Quasi-Monte Carlo (QMC) methods reveals a smaller estimator variance and faster convergence rate of the latter method in this study.
Robust Estimation of Scatter Matrix, Random Matrix Theory and an Application to Spectrum Sensing(2019-05-05) [Thesis]
Advisor: Alouini, Mohamed-Slim
Committee members: Rue, Haavard; Kammoun, AblaThe covariance estimation is one of the most critical tasks in multivariate statistical analysis. In many applications, reliable estimation of the covariance matrix, or scatter matrix in general, is required. The performance of the classical maximum likelihood method relies a great deal on the validity of the model assumption. Since the assumptions are often approximately correct, many robust statistical methods have been proposed to be robust against the deviation from the model assumptions. M-estimator is an important class of robust estimator of the scatter matrix. The properties of these robust estimators under high dimensional setting, which means the number of dimensions has the same order of magnitude as the number of observations, is desirable. To study these, random matrix theory is a very important tool. With high dimensional properties of robust estimators, we introduced a new method for blind spectrum sensing in cognitive radio networks.
Modular Autonomous Taxiing Simulation and 3D Siamese Vehicle Tracking(2019-05) [Thesis]
Advisor: Ghanem, Bernard
Committee members: Al-Naffouri, Tareq Y.; Thabet, Ali KassemThe automation of navigation for different kinds of vehicles is a research problem of great interest. This problem has applications with unmanned aerial vehicles (UAVs) as well as manned vehicles such as cars and planes. The goal of an autonomous vehicle is to navigate safely from one point to another given a set of high-level instructions and data from a set of sensors. This thesis explores an implementation of a modular approach for autonomously driving taxiing planes before proposing methods for object tracking using a LIDAR sensor which can be incorporated into the autonomous driving pipeline. The taxiing algorithm regresses waypoints for the plane to follow given a high-level driving goal such as ”turn left” or ”go straight”, along with RGB images taken from the cockpit and wings. Waypoints are then used with a separate control system to taxi the plane. The training and testing of this autonomous aircraft is done in a photo-realistic simulator which has been adapted for this task. The policy developed in this fashion is capable of learning how to go straight and how to turn. However, the driving policy is not trained to react to other moving objects. To address this issue, and due to the superior reliability of LIDAR over RGB sensors, an object tracking method using only LIDAR point clouds is proposed. The proposed method uses a novel 3D Siamese network to obtain a similarity score between a model and candidate object point clouds. This similarity score is shown to work for tracking by applying it using an exhaustive search and obtaining improved performances when compared with simple baselines. For a realistic application, the similarity score is applied using candidates provided by a search on the BEV projection of the LIDAR point cloud. This method is shown to provide improved tracking results over other search strategies when using a lower number of candidates.
Gain Enhancement Techniques for mm-wave On-chip Antenna on Lossy CMOS Platforms(2019-05) [Thesis]
Advisor: Shamim, Atif
Committee members: Bagci, Hakan; Wu, YingRecently, there is great interest in achieving higher-level integration, higher data rates, and reduced overall costs. At millimeter-wave (mm-wave) bands, the wavelength is small enough to realize an antenna-on-chip (AoC), which is an ideal solution for high compactness and lower costs. However, the main drawback of AoC is the low resistivity (10 Ω-cm) Si substrate used in the standard CMOS technology, which absorbs most radio-frequency (RF) power that was supposed to be radiated by the on-chip antenna. Moreover, due to the high relative permittivity (11.9) and relatively large electrical thickness of the Si, higher order surface wave modes get excited, which further degrade the antenna radiation performance. In order to alleviate the above-mentioned issues with the low gain of AoC, a combination of an artificial magnetic conductor (AMC) surface, a high dielectric constant superstrate, and a Fresnel lens is presented in this work. The AMC is realized in standard CMOS technology along with the AoC, whereas the superstrate and lens are part of a smart packaging solution. The AMC surface can change wave propagation characteristics at the operating frequency to achieve in-phase reflection, resulting in gain enhancement by reducing the loss in the substrate. The high dielectric constant superstrate behaves as an impedance transformer between the Si substrate and air, thus enhancing the coupling to air. Finally, the Fresnel lens enhances the gain by focusing the electromagnetic (EM) radiation beam at the boresight. For AoC realization, a standard 0.18 μm CMOS process was utilized. A coplanar waveguide (CPW) fed monopole on-chip antenna at 71 GHz, along with the corresponding driving circuit, was designed and fabricated. The AMC enhances the gain by 3 dB. Since the chip needs to be packaged anyways, in this work, we optimize the package to provide further gain enhancement. This smart package, comprising a superstrate and a Fresnel lens, provides a gain enhancement of 16 dB. The overall combination of the optimized AMC surface, superstrate layer, and lens package can provide a gain enhancement of around 19 dB. Furthermore, the package has been realized through additive manufacturing techniques that ensure lower costs for the overall system.
Model Predictive Control and State Estimation for Membrane-based Water Systems(2019-05) [Thesis]
Advisor: Laleg-Kirati, Taous-Meriem
Committee members: Shamma, Jeff S.; Hong, Pei-YingLack of clean fresh water is one of the most pervasive problems afflicting people throughout the world. Efficient desalination of sea and brackish water and safe reuse of wastewater become an insistent need. However, such techniques are energy intensive, and thus, a good control design is needed to increase the process efficiency and maintain water production costs at an acceptable level. This thesis proposes solutions to the above challenges and in particular will be focused on two membranebased water systems: Membrane Distillation (MD) and Membrane Bioreactor (MBR) for wastewater treatment plant (WWPT). The first part of this thesis, Direct Contact Membrane Distillation (DCMD) will study as an example an MD process. MD is an emerging sustainable desalination technique which can be powered by renewable energy. Its main drawback is the low water production rate. However, it can be improved by utilizing advanced control strategies. DCMD is modeled by a set of Differential Algebraic Equations (DAEs). In order to improve its water production, an optimization-based control scheme termed Model Predictive Control (MPC) provides a natural framework to optimally operate DCMD processes due to its unique control advantages. Among these advantages are the flexibility provided in formulating the objective function, the capability to directly handle process constraints, and the ability to work with various classes of nonlinear systems. Motivated by the above considerations, two MPC schemes that can maximize the water production rate of DCMD systems have been developed. The first MPC scheme is formulated to track an optimal set-point while taking input and stability constraints into account. The second MPC scheme, Economic MPC (EMPC), is formulated to maximize the distilled water flux while meeting input, stability and other process operational constraints. The total water production under both control designs is compared to illustrate the effectiveness of the two proposed control paradigms. Simulation results show that the DCMD process produces more distilled water when it is operated by EMPC than when it is operated by MPC. The above control techniques assume the full access to the system states. However, this is not the case for the DCMD plant. To effectively control the closed-loop system, an observer design that can estimate the values of the unmeasurable states is required. Motivated by that, a nonlinear observer design for DCMD is proposed. In addition, the effect of the estimation gain matrix on the differentiation index of the DAE system is investigated. Numerical simulations are presented to illustrate the effectiveness of the proposed observer design. The observer-based MPC and EMPC are also studied in this work. Mathematical modeling of a wastewater treatment system is critical because it enhances the process understanding and can be used for process design and process optimization. Motivated by the above considerations, modeling and optimal control strategies have been developed and applied to the MBR-based wastewater treatment process. The model is an extension of the well-known Benchmark simulation models for wastewater treatment. In addition, model predictive control has been applied to maintain the dissolved oxygen concentration level at the desired value. In addition, a conventional PID controller has also been developed. The simulation results show that the both of controllers can be used for dissolved oxygen concentration control. However, MPC has better performance compared to PID scenario.
Expandable Polymer Assisted Wearable Personalized Medicinal Platform(2019-05) [Thesis]
Advisor: Hussain, Muhammad Mustafa
Committee members: Alouini, Mohamed-Slim; Schwingenschlögl, UdoConventional healthcare and the practice of medicine largely relies on the ineffective concept of one size fits all. Personalized medicine is an emerging therapeutic approach that aims to develop an advanced therapeutic technique that provides tailor-made therapy based on every individuals’ needs by delivering the right drug at the right time with the right amount of dosage. The advancement in technologies such as flexible electronics, microfluidics, biosensors, and advanced artificial intelligence can enable the realization of a truly effective personalized therapy. However, currently, there is a lack for a personalized minimally-invasive wearable closed-loop drug delivery system that is continuous, automated, conformal to the skin and cost-effective. Thus, this thesis focuses on the design, fabrication, optimization, and application of an automated personalized microfluidics drug delivery platform augmented with flexible biosensors, heaters, and expandable polymeric actuator. The platform provides precise drug delivery with spatiotemporal control over the administered dose as a response to real-time physiological changes of the individual. The system is flexible enough to be conformal to the skin and drug is transdermally administered through biocompatible microneedles. The platform includes a flexible multi-reservoir microfluidics layer, flexible and conformal heating elements, skin sensors and processing units which are powered by a lightweight battery integrated into the platform. The developed platform was fabricated using rapid, cost-effective techniques that are independent of advanced microfabrication facilities to expand its applications to low-resource setting and environments.
Resource Allocation in Future Terahertz Networks(2019-05) [Thesis]
Advisor: Alouini, Mohamed-Slim
Committee members: Shihada, Basem; Amin, OsamaTerahertz (THz) band represents the unused frequency band between the microwave and optical bands and lies in the range of frequencies between 0.1 to 10 THz. As a result, the THz signal generation can be done using electronic or photonic circuits. Moreover, the channel gain has hybrid features from both microwave and optical bands allowing to reap the benefits of each band. Adopting such a technology can mitigate the spectrum scarcity and introduce a substantial solution to other systems such as visible light communications. Despite of the generous bandwidth, the THz communications suffer from high attenuation that increases with adopted frequency similar to the microwave frequency band. Furthermore, THz communications are subject to a different type of attenuation called Molecular Absorption, that depends on the chemical nature of the ambiance air. Thus, THz transmitters need to use extra power and high antenna gains to overcome signal loss and compensate the short distance range limitation. In this thesis, we investigate the pathloss model to compute the overall attenuation faced by the THz wave for different frequencies and weather conditions. Then, we use the THz technology to support the operation of uplink networks using directional narrow beams. We optimize the uplink communication network resource represented in the frequency bands and the assigned power in order to minimize the total power consumption while achieving a specific quality of service. Furthermore, we investigate the impact of weather conditions and the system’s requirements in order to guarantee a better performance.
Towards Multistate Magnetic Tunnel Junctions for Memory and Logic Applications(2019-05) [Thesis]
Advisor: Fariborzi, Hossein
Committee members: Shamim, Atif; Kosel, JürgenFor many decades, the revolution in semiconductor industry has continuously been powered by the successful down scaling of complementary metal-oxide semiconductor (CMOS) technology to produce integrated circuits with improved performance at lower cost. However, current charge-based CMOS technology is already approaching physical limits and, thus, encounters a number of technological challenges. Spintronics is an emerging and rapidly evolving research field that has a great potential to overcome these challenges confronting CMOS by introducing the electron spin, in addition to electron charge, as an extra degree of freedom. Traditional spintronic devices are based on the alignment of spins in magnetic layers, manipulated by spin-polarized currents. Thus, employing the non-volatile nature of layer magnetization and its direction to represent the bit state, spintronics provides power-efficient devices that are attractive for memory and logic applications. Magnetoresistive random access memory (MRAM) is one of the most essential applications of spin based electronics, which has already been recognized as the leading candidate for future universal memory. MRAM cells use spin-based magnetic tunnel junctions (MTJs) as the fundamental storage blocks. These conventional MTJs employ the use of magnetic elements with a single axis of magnetization, which provide two resistance states, capable of storing one bit of information. Enhancing the memory density is one of the major challenges encountered by MRAM industry, as the straightforward approach of reducing the magnetic bit size is unfeasible with magnetic devices due to intrinsic superparamagnetism effects. In this thesis, we propose increasing the bit density in MRAM by implementing shape anisotropy induced multistate MTJs. By patterning the free ferromagnetic layer of MTJs in the shape of four intersecting ellipses we achieve four in-plane stable axes of magnetization, capable of providing eight resistance states in total, the switching between which is performed by spin-orbit torques (SOT) in spin Hall metals (SHM). We initially verify the proposed concept with micromagnetic simulations followed by fabrication and, consequent, room temperature characterization of the first experimental prototypes.
Optimization of Process Variables for Oxidative Coupling of Methane(2019-05) [Thesis]
Advisor: Sarathy, Mani
Committee members: Gascon, Jorge; Farooq, AamirOxidative coupling of methane (OCM) is a promising route for converting abundant natural gas resources into more useful chemicals like paraffins and olefins (primarily C2). However, to date, there is no current OCM production plant due to low overall conversion and selectivity to the desired product(s). In this work, different operating factors are studied experimentally and through simulation with respect mainly to three responses: CH4 conversion, C2 main product selectivity, and COx side product selectivity. The aim is to identify the best operating condition for maximum ethylene production combined with COx production. Design of experiments (DoE) method was used to analyze the experimental results by applying the full factorial approach. Simulation results were studied by finding the correlation strength between input factors and responses. It was found that the performance of an OCM reactor could be greatly improved under optimal operating conditions. Operating temperature and CH4/O2 ratio have the highest influence while catalyst weight and flow rate have the lowest influence on the OCM responses and mainly depend on rector dimensions.
Structural Analysis of Arabidopsis thaliana CDC48A ATPase using Single Particle Cryo-Electron Microscopy(2019-05) [Thesis]
Advisor: Arold, Stefan T.
Committee members: Jaremko, Łukasz; Gao, XinCdc48A and its human homologue P97 are from ATPase family, which play a variety of roles in cellular activates and it has a crucial involvement in protein quality control pathways. It is best known for its involvement in endoplasmic reticulum associated protein degradation (ERAD), where it mediates the degradation of the aggerated or misfolded proteins by the proteasome. Considering the multiple functions of Cdc48A in many protein regulatory processes, it is a potential therapeutic target for neurogenerative diseases and cancer. Cdc48A polypeptide comprises N domain, followed by D1 and D2 domains respectively that are joined by linkers, whereas functionally it forms a homo hexameric complex. Since Cdc48A is from the ATPase family, it uses the ATP hydrolysis to generate a mechanical force with its co-factors to perform its functions. There are many cofactors that interact with Cdc48A and two of them are Ufd1-NpI4 which in turn interact with ubiquitinated proteins from the ER membrane. The mechanism linking the conversion of the energy of ATP hydrolysis into mechanical force and unfolding the substrate is vague. My aim is to reconstruct a first 3D- model of plant Cdc48A using single particle cryo-EM, which serves the basis to conduct more detailed mechanistic studies towards substrate unfolding and threading/unfolding in the future. In general, results showed one defined structure of cdc48A at ~ 9.8 Å, which is the ADP-ATP conformation. Although another other structure was also resolved at ~ 8.9 Å, it was hard to characterize due to its dissimilarity with known structures of Cdc48A homologues and thus requires further characterization.
Privacy Concerned D2D-Assisted Delay-Tolerant Content Distribution System(2019-04-28) [Thesis]
Advisor: Shihada, Basem
Committee members: Alouini, Mohamed-Slim; Amin, OsamaIt is foreseeable that device-to-device (D2D) communication will become a standard feature in the future, for the reason that it offloads the data traffic from network infrastructures to user devices. Recent researches prove that delivering delay-tolerant contents through content delivery network (CDN) by D2D helps network operators increase spectral and energy efficiency. However, protecting the private information of mobile users in D2D assistant CDN is the primary concern, which directly affects the willingness of mobile users to share their resources with others. In this thesis, we proposed a privacy concerned top layer system for selecting the sub-optimal set of mobile nodes as initial mobile content provider (MCP) for content delivery in any general D2D communications, which implies that our proposed system does not rely on private user information such as location, affinity, and personal preferences. We model the initial content carrier set problem as an incentive maximization problem to optimize the rewards for network operators and content providers. Then, we utilized the Markov random field (MRF) theory to build a probabilistic graphical model to make an inference on the observation of delivered contents. Furthermore, we proposed a greedy algorithm to solve the non-linear binary integer programming (NLBIP) problem for selecting the optimal initial content carrier set. The evaluations of the proposed system are based on both a simulated dataset and a real-world collected dataset corresponding to the off-line and on-line scenarios.
Theoretical and Experimental Studies of Optical Properties of BAlN and BGaN Alloys(2019-04-21) [Thesis]
Advisor: Li, Xiaohang
Committee members: Schwingenschlögl, Udo; Ooi, Boon S.Wurtzite III-nitride semiconductor materials have many technically important applications in optical and electronic devices. As GaN-based visible light-emitting diodes (LEDs) and lasers starts to mature, interest in developing UV devices starts to rise. The search for materials with larger bandgaps and high refractive index contrast in the UV range has inspired multiple studies of BN-based materials and their alloys with traditional III-nitrides. Additionally, alloying III-nitrides with boron can reduce their lattice parameters giving a new option for strain engineering and lattice matching. In this work I investigate the refractive indices of BAlN and BGaN over the entire compositional range using hybrid density functional theory (DFT). An interesting non-linear trend of the refractive index curves appears as boron content is increased in the BAlN and BGaN alloys. The results of this calculation were interpolated and plotted in three dimensions for better visualization. This interpolation gives a 3D dataset that can be used in designing a myriad of devices at all binary and ternary alloy compositions in the BAlGaN system. The interpolated surface was used to find an optimum design for a strain-free, high reflection coefficient and high bandwidth DBR. The performance of this DBR was quantitatively evaluated using finite element simulations. I found that the maximum DBR reflectivity with widest bandwidth for our materials occurs at a lattice parameter of 3.113 Å using the generated 3D dataset. I use the corresponding material pair to simulate a DBR at the wavelength 375 nm in the UVA range. A design with 25 pairs was found to have a peak reflectivity of 99.8%. This design has a predicted bandwidth of 26 nm measured at 90% peak performance. The high reflectivity and wide bandwidth of this lattice-matched design are optimal for UVA VCSEL applications. I have assisted in exploring different metalorganic chemical vapor deposition (MOCVD) techniques, continuous growth and pulsed-flow modulation, to grow and characterize BAlN alloys. Samples grown using continuous flow show better optical quality and are characterized using spectroscopic ellipsometry. The refractive index of samples obtained experimentally is significantly below the predicted value using DFT.
Power Adaption Over Fluctuating Two-Ray Fading Channels and Fisher-Snedecor F Fading Channels(2019-04) [Thesis]
Advisor: Alouini, Mohamed-Slim
Committee members: Al-Naffouri, Tareq Y.; Park, Ki-HongIn this thesis, we investigate the ergodic capacity under several power adaption schemes, including optimal power and rate algorithm (OPRA), optimal rate algo rithm (ORA), channel inversion (CI), and truncated channel inversion (TCI), over fluctuating two-ray (FTR) fading channels and Fisher-Snedecor F fading channels. After some mathematical manipulations, the exact expressions for the EC under those power adaption schemes are derived. To simplify the expressions and also get some insights from the analysis, the corresponding asymptotic expressions for the EC are also derived in order to show the slope and power offset of the EC in the high signal-to-noise ratio (SNR) region. These two metrics, i.e., slope and power offset, govern the EC behaviour in the high SNR region. Specifically, from the derived asymptotic expressions, we find that the slope of the EC of OPRA and ORA over FTR fading channels is always unity with respect to the average SNR in the log-scale in high SNRs, while the asymptotic EC of the TCI method is not a line function in the log-scale. For the Fisher-Snedecor F fading channel, the slope of asymptotic EC under OPRA, ORA, and CI (m > 1) schemes is unity in the log-scale, where m is the fading parameter. The slope of the TCI method depends on m, i.e., unity for m > 1 and m for m > 1, while the asymptotic EC of TCI is not a line function for m = 1. Finally, Monte-Carlo simulations are used to demonstrate the correctness of the derived expressions.
Wavelength Dependence of Underwater Turbulence Characterized Using Laser-Based White Light(2019-04) [Thesis]
Advisor: Ooi, Boon S.
Committee members: Alouini, Mohamed-Slim; Shihada, Basem; Ng, Tien KheeThe means of communication in oceanic environments is currently dominated by sonar. Although it is reliable for long-distance transmission, the vision of internet of underwater things (IoUT) requires an alternate means for high-data-rate transmission. It is also envisaged that a networked underwater and above-water objects, such as sensor nodes, and autonomous underwater vehicles will benefit seafloor exploration. The use of laser-based optical communication is poised to realize this dream while working hand-in-hand with acoustic and radio-frequency technologies from the littoral zone to deep blue sea. While blue and green lasers are typically utilized depending on the optical properties of the water, laser-based white light is attractive in a number of aspects. In this thesis, we proposed and realized the use of white light to model the channel and to provide the immediate decision for the preferred system configuration, which is critical for developing reliable communication links, particularly, in the presence of turbulence, which makes the alignment of underwater wireless optical communication (UWOC) links challenging. Temperature and salinity changes are among factors that change the refraction index, giving rise to beam wander. This thesis explores the dependence of underwater turbulence on the wavelength. After comparing the performance of red, green, and blue lasers, an ultra-fast comprehensive method that utilizes a white-light source that can produce a wide range of wavelengths is implemented. Experimental results show an 80%-decrease in the scintillation index as the wavelength is increased from 480 to 680 nm in weak turbulence caused by a 0.02-℃/cm temperature gradient with a 40-ppt salt concentration, which emulates conditions found in the Red Sea. The effect of turbulence on the bit error ratio (BER) is also investigated experimentally. Temperature gradients increased the BER especially for shorter wavelengths. The results along long-transmission distances were verified using Monte Carlo simulations. The correlation matrix between wavelengths was studied, which is important for designing multiple-input multiple-output systems. The results obtained show that as the difference in the wavelengths increases, the correlation decreases. Based on the interplay among scintillations, scattering, absorption, and the correlation between different wavelengths, it is possible to design a more reliable UWOC link.
Object Detection Using Multiple Level Annotations(2019-04) [Thesis]
Advisor: Ghanem, Bernard
Committee members: Al-Naffouri, Tareq Y.; Thabet, Ali KassemObject detection is a fundamental problem in computer vision. Impressive results have been achieved on large-scale detection benchmarks by fully-supervised object detection (FSOD) methods. However, FSOD approaches require tremendous instance-level annotations, which are time-consuming to collect. In contrast, weakly supervised object detection (WSOD) exploits easily-collected image-level labels while it suffers from relatively inferior detection performance. This thesis studies hybrid learning methods on the object detection problems. We intend to train an object detector from a dataset where both instance-level and image-level labels are employed. Extensive experiments on the challenging PASCAL VOC 2007 and 2012 benchmarks strongly demonstrate the effectiveness of our method, which gives a trade-off between collecting fewer annotations and building a more accurate object detector. Our method is also a strong baseline bridging the wide gap between FSOD and WSOD performances. Based on the hybrid learning framework, we further study the problem of object detection from a novel perspective in which the annotation budget constraints are taken into consideration. When provided with a fixed budget, we propose a strategy for building a diverse and informative dataset that can be used to optimally train a robust detector. We investigate both optimization and learning-based methods to sample which images to annotate and which level of annotations (strongly or weakly supervised) to annotate them with. By combining an optimal image/annotation selection scheme with the hybrid supervised learning, we show that one can achieve the performance of a strongly supervised detector on PASCAL-VOC 2007 while saving 12:8% of its original annotation budget. Furthermore, when 100% of the budget is used, it surpasses this performance by 2:0 mAP percentage points.
Numerical study of linear and nonlinear problems using two-fluid plasma model in one dimension(2019-04) [Thesis]
Advisor: Samtaney, Ravi
Committee members: Farooq, Aamir; Wu, YingThe ideal two-fluid plasma model is a more generalized plasma model compared to the ideal MHD and it couples the ion and electron Euler equations via Maxwell's equations. Two-fluid plasma model is essential when the ion and electron fluids are at different temperatures. In this work, a fundamental investigation on the effect of non-dimensional light speed, ion-to-electron mass ratio and plasma beta on the plasma dynamics in the Brio-Wu shock tube Riemann problem is presented. A one dimensional finite volume code is developed based on the macroscopic governing equations, with second order accuracy in space and time. The source terms are treated implicitly and the homogeneous flux terms are treated explicitly. The credibility of the numerical results is assessed by performing several linear and nonlinear tests. Realistic light speed results in increasing the stiffness of the equations and severe time step restriction, which poses a challenge to the numerical simulations. In the context of the Brio-Wu shock tube problem, it is observed that the light speed is not important with respect to the hydrodynamics. However, light speed does affect the magnitude of the self generated electric field. Mass ratio affects the electron plasma dynamics. The speed of the fast moving electron plasma waves changes with the mass ratio. The results obtained using a mass ratio of 500 are in close agreement with that of realistic mass ratio of 1836. Increasing plasma beta suppresses the amplitude of the fast moving electron plasma waves.
Standardized short-term bleaching assays resolve differences in coral thermotolerance across microhabitat reef sites(2019-04) [Thesis]
Advisor: Voolstra, Christian R.
Committee members: Aranda, Manuel; Tester, Mark A.Coral bleaching is now the main driver of reef degradation. The common notion is that most corals bleach and suffer mortality at just 1-2°C above their mean summer maximum temperatures, but some species and genotypes resist or recover better than others. Here we conducted a series of 18-hour short-term heat stress assays side-by-side with a long-term heat stress experiment to assess the ability of both approaches to resolve putative differences in coral thermotolerance and provide a measure of in situ reef temperature thresholds. Using a suite of measures (photosynthetic performance, coral whitening, chlorophyll a, host protein, algal symbiont counts, and algal type association), we assessed bleaching sensitivity/resilience of Stylophora pistillata colonies from the exposed and protected sides of a near-shore coral reef in the central Red Sea. As suggested by the differential mortality during a previous bleaching event, coral colonies from the protected site exhibited less impacted physiological performance in comparison to their exposed site counterparts, and these differences were resolved using both experimental setups. Notably, the long-term experiment provided better resolution with regard to the different measures collected, but at the price of portability, cost, and duration of the experiment. Variability in resilience to ocean warming is critical to reef persistence, yet we lack standardized diagnostics to rapidly assess bleaching severity or resilience across different corals and locations. Using a newly developed portable experimental system termed CBASS (the Coral Bleaching Automated Stress System), we demonstrate that mobile, short-term heat stress assays can resolve fine-scale differences in coral thermotolerance across reef sites. Based on our results, photosynthetic efficiency measured by non-invasive PAM fluorometry provides a rapid and representative proxy of coral resilience. Our system holds the potential to be employed for large-scale determination of in situ bleaching temperature thresholds across reef sites and species. Such data can then be used to identify resistant genotypes (and reefs) for downstream experimental examination and to complement existing remote-sensing approaches.
High-Performance Polyimide Gas Separation Membranes Based on Triptycene Dianhydrides and Di-Hydroxy-Diamino-Triptycene Monomers.(2019-04) [Thesis]
Advisor: Pinnau, Ingo
Committee members: Lai, Zhiping; Han, YuDistillation technology involves capital- and energy-intensive processes for light olefin/paraffin separation. Global demand for propylene has already exceeded 110 million tons per year. Therefore, distillation processes used for the separation of C3H6/C3H8 should be replaced or debottlenecked with more efficient and cost-effective technology. In the last three decades, membrane-based gas separation processes have successfully emerged, thus competing with conventional separation processes. Membranes potentially offer lower capital investment and operation cost than distillation columns. In this study, the use of advanced membrane materials for C3H6/C3H8 separation was investigated. Three novel triptycene-based polyimides were synthesized by Dr. Bader Ghanem from one diamine monomer, namely 2,6-dihydroxy-3,7-diaminotriptycene (DTA1-OH), and three dianhydride monomers, (i) non-substituted triptycene tetracarboxylic dianhydride (TDA), (ii) 9,10-dimethyltriptycene tetracarboxylic dianhydride (TDA1) and (iii) 9,10-iso-propyltriptycene tetracarboxylic dianhydride (TDAi3). It is important to note that polyimide membranes based on triptycene dianhydrides and triptycene diamines have never been reported in the literature before. Pure-gas permeability coefficients of He, H2, N2, O2, CO2, CH4, C3H6, and C3H8 were determined at 2 bar and 35 °C. Furthermore, C3H6 and C3H8 gas sorption isotherms were measured by gravimetric techniques, and experimental data were collected up to 7 bar at 35 °C. TDA-DAT1-OH, TDA1-DAT1-OH, TDAi3-DAT1-OH exhibited C3H6 permeability of 12.1, 16.6, and 5.64 Barrer with pure-gas C3H6/C3H8 selectivity of 35.7, 29.6, and 32.8 respectively. These properties exceeded the 2003 pure-gas upper bound for C3H6/C3H8. The BET surface area increased in the order of TDA-DAT1-OH (437 m2/g) < TDAi3-DAT1-OH (467 m2/g) < TDA1-DAT1-OH (557 m2/g). The frecational free volume (FFV) increased in the order of TDAi3-DAT1-OH (0.25) < TDA-DAT1-OH (0.28) < TDA1-DAT1-OH (0.30). TDA1-DAT1-OH (109 μm) showed less and slower physical aging than TDA-DAT1-OH (94 μm) after 60 days, where the O2 and CO2 permeability of both polyimides decreased by about 40% and 69%, respectively. After 30 days, TDAi3-DAT1-OH displayed the highest selectivity gain relative to its counterparts and exceeded the 2008 upper bound for CO2/CH4. TDA1-DAT1-OH exhibited 7-fold higher C3H6 permeability coupled with almost 3-fold higher C3H6/C3H8 selectivity relative to a previously reported commercial polyphenylene oxide (PPO) membrane.
Applications in computational structural biology: the generation of a protein modelling pipeline and the structural analysis of patient-derived mutations(2019-04) [Thesis]
Advisor: Arold, Stefan T.
Committee members: Gao, Xin; Jaremko, ŁukaszBesides helping us advance the understanding of the physicochemical principles governing the three-dimensional folding of proteins and their mechanisms of action, the ability to build, evaluate, and optimize reliable 3D protein models has provided valuable tools for the development of different applications in the fields of biotechnology, medicine, and synthetic biology. The development of automated algorithms has made many of the current methodologies for protein modelling and visualization available to researchers from all backgrounds, without the need to be familiarized with the inner workings of their statistical and biophysical principles. However, there is still a lack in some areas where the learning curves are too steep for the methods to be widely used by the average non-programmer molecular biologist, or the implementation of the methods lacks key features to improve the interpretability and impact of their results. Throughout this work, I will focus on two different applications in the field of structural biology where computational methods provide useful tools to aid in synthetic biology or medical research. The first application is the implementation of a pipeline to build models of protein complexes by joining structured domains with disordered linkers, in individual or multiple chains, and with the possibility of building symmetric structures. Its capabilities and performance for the generation of complex constructs are evaluated, and possible areas of improvement described. The second application, but not less important, involves the structural analysis of patient-derived protein mutants using protein modelling techniques and visualization tools, to elucidate the potential molecular basis for the patient’s phenotype. The methodology for these analyses is described, along with the results and observations from 22 such cases in 13 different proteins. Finally, the need for a dedicated pipeline for the structure-based prediction of the effect of different types of mutations on the stability and function of proteins, complementary to available sequence-based approaches, is highlighted.