Engineering Commons^™

Approximation And Relaxation Approaches For Parallel And Distributed Machine Learning, Stephen Tyree

McKelvey School of Engineering Theses & Dissertations

Large scale machine learning requires tradeoffs. Commonly this tradeoff has led practitioners to choose simpler, less powerful models, e.g. linear models, in order to process more training examples in a limited time. In this work, we introduce parallelism to the training of non-linear models by leveraging a different tradeoff--approximation. We demonstrate various techniques by which non-linear models can be made amenable to larger data sets and significantly more training parallelism by strategically introducing approximation in certain optimization steps.

For gradient boosted regression tree ensembles, we replace precise selection of tree splits with a coarse-grained, approximate split selection, yielding both faster …

Accelerating Heuristic Search For Ai Planning, You Xu

McKelvey School of Engineering Theses & Dissertations

AI Planning is an important research field. Heuristic search is the most commonly used method in solving planning problems. Despite recent advances in improving the quality of heuristics and devising better search strategies, the high computational cost of heuristic search remains a barrier that severely limits its application to real world problems. In this dissertation, we propose theories, algorithms and systems to accelerate heuristic search for AI planning.

We make four major contributions in this dissertation. First, we propose a state-space reduction method called Stratified Planning to accelerate heuristic search. Stratified Planning can be combined with any heuristic search to …

A Four-Dimensional Image Reconstruction Framework For Pet Under Arbitrary Geometries, Aswin John Mathews

McKelvey School of Engineering Theses & Dissertations

Positron Emission Tomography (PET) is a functional imaging modality with applications ranging from the treatment of cancer, studying neurological diseases and disease models. Virtual-Pinhole PET technology improves the image quality in terms of resolution and contrast recovery. The technology calls for having a detector with smaller crystals placed near a region of interest in a conventional whole-body PET scanner. The improvement is from the higher spatial sampling of the imaging area near the detector. A prototype half-ring PET insert built to study head-and-neck cancer imaging was extended to breast cancer imaging. We have built a prototype half-ring PET insert for …

Aerosol Techniques For Deposition And Characterization Of Biological And Biomimetic Sensitizers For Solar Devices, Vivek B. Shah

McKelvey School of Engineering Theses & Dissertations

Nano-structured columnar films synthesized by the aerosol chemical vapor deposition (ACVD) system are unique, and have proved to be useful for fabricating solar cells and in other applications. The film formation by ACVD process involves three main phenomenon - aerosol formation and growth, deposition, and restructuring. In this work, thin film formation by the ACVD process is simulated by combining three models - (A) particle formation in gas phase by atrimodal model, (B) particle deposition onto heated substrate by a Brownian dynamics model and (C) sintering on the heated substrate by a multi-particle geometric sintering model (MPGSM). Modelling and simulation …

Evaluation Of The Performance Of Various Turbulence Models For Accurate Numerical Simulation Of A 2d Slot Nozzle Ejector, Colin Graham

McKelvey School of Engineering Theses & Dissertations

With the development over the last several decades, accurate Computational Fluid Dynamics (CFD) modeling has now become an essential part in the analysis and design of various industrial products where the fluid flow plays an important role. The goal of this thesis is to apply the CFD technology to the analysis of a 2D slot nozzle ejector which has application in Short Take-off and Landing (STOL) aircraft and other future aerospace vehicles. In the nozzle-ejector configuration, the high speed air flow from the nozzle entrains the ambient air into a mixing chamber (ejector) as a means to create additional thrust …

Shape Optimization Of Airfoils Without And With Ground Effect Using A Multi-Objective Genetic Algorithm, Yilei He

McKelvey School of Engineering Theses & Dissertations

The focus of this thesis is on shape optimization using a genetic algorithm. A multi-objective genetic algorithm (MOGA) is employed to optimize the shape of an airfoil to improve its lift and drag characteristics, in particular to achieve two objectives simultaneously that is to increase its lift as well as its lift to drag ratio. The commercially available software FLUENT is employed to calculate the flow field on an adaptive structured mesh, which is generated by the commercial mesh generating software ICEM. The flow field is calculated using the Reynolds-Averaged Navier-Stokes (RANS) equations in conjunction with a two equation k-ω …

Engineering Biosensors For Short-Chain Alcohols, Yu Xia

McKelvey School of Engineering Theses & Dissertations

Biofuel is a promising substitute for fossil fuel and the research of biofuel production has been extensively conducted during the recent years. Great efforts have been made to create many types of the fuel production hosts. However, effective approaches for high specificity and high throughput screening of the fuel production strains are still lacking. The cellular stress response is one universal defense mechanism when a microbial cell is exposed to an unfavorable substance or environment. It triggers a series of downstream responses when a cell senses certain chemicals. Transcriptional factors are widely used in living organisms to regulate gene expression. …

Single Cuo Nanowires Electrical Properties And Application On Photoelectrochemical Water Splitting, Junnan Wu

McKelvey School of Engineering Theses & Dissertations

Charge transport in single crystal, p type copper oxide (CuO) nanowires (NWs) was studied through temperature based (120 K – 400 K) current-voltage (I-V) measurements of 2-terminal single NW devices. Individual CuO NWs with an average diameter of 220 nm were attached to Au electrodes 2.5 µm apart, using a dielectrophoresis method. I-V curves showed a transition from linear behavior at low bias to strong power law dependence (I ~ V^a) at high bias, which can be attributed to space charge limited current (SCLC) mechanism. At low electrical fields (< 0.89 × 10³ V·cm^-1), the number of …

Optical Resonators And Fiber Tapers As Transducers For Detection Of Nanoparticles And Bio-Molecules, Huzeyfe Yilmaz

McKelvey School of Engineering Theses & Dissertations

In recent years, detection of biological interactions on single molecule level has aspired many researchers to investigate several optical, chemical, electrical and mechanical sensing tools. Among these tools, toroidal optical resonators lead the way in detection of the smallest particle/molecule with the real time measurements. In this work, bio-sensing capabilities of toroidal optical resonators are investigated. Bio-sensing is realized via measuring the analyte-antigen interaction while the antigen is immobilized through a novel functionalization method.

Not long ago, detection of single nanoparticles using optical resonators has been accomplished however the need for cost-effective and practical transducers demands simpler tools. A tapered …

Detection Of Parkinson Disease Rest Tremor, Matthew J. Johnson

McKelvey School of Engineering Theses & Dissertations

Parkinson Disease (PD) is a debilitating and progressive movement disorder that is estimated to affect over six million worldwide. One of the most characteristic symptoms of PD is resting tremor, which involves unintentional and rhythmic muscle oscillations of an afflicted extremity while the muscles of said extremity are relaxed. This study involved measuring the rest tremor of 10 PD subjects, 10 Essential Tremor subjects, and 10 healthy control subjects using two devices. One device was an FDA approved accelerometry system to measure human tremor known as the Tremorometer^TM and the other was a consumer three-dimensional camera known as the …

Engineered Plasmonic Nanostructures For Nanomedicine, Christopher Matthew Portz

McKelvey School of Engineering Theses & Dissertations

No abstract provided.

Numerical Simulation And Optimization Of Carbon Dioxide Utilization And Storage In Enhanced Gas Recovery And Enhanced Geothermal Systems, James H. Biagi

McKelvey School of Engineering Theses & Dissertations

With rising concerns surrounding CO₂ emissions from fossil fuel power plants, there has been a strong emphasis on the development of safe and economical Carbon Capture Utilization and Storage (CCUS) technology. Two methods that show the most promise are Enhanced Gas Recovery (EGR) and Enhanced Geothermal Systems (EGS). In Enhanced Gas Recovery a depleted or depleting natural gas reservoir is re-energized with high pressure CO₂ to increase the recovery factor of the gas. As an additional benefit following the extraction of natural gas, the reservoir would serve as a long-term storage vessel for the captured carbon. CO_{2 …}

An Approach To Thermocouple Temperature Measurements That Reduces Uncertainties Associated With Radiative Corrections, Siddharth Krishnan

McKelvey School of Engineering Theses & Dissertations

Obtaining accurate temperature measurements in flame environments with thermocouples is extremely challenging due to the effects of radiative heat loss. These losses are difficult to quantify and often cannot be corrected for or minimized without sacrificing spatial resolution. In this work, a new experimental methodology is presented that has shown potential to minimize the temperature correction by both increasing and controlling the effects of convection. This is accomplished through high speed rotation of the thermocouple. The rotation yields a high and known convective velocity over the thermocouple. Heat transfer can then be modeled for the thermocouple, and a functional relationship …

Real-Time Image Processing On An Fpga For An Intraoperative Goggle Device, Yiyi Zhang

McKelvey School of Engineering Theses & Dissertations

The emergence of near-infrared dyes for fluorescence imaging has had a tremendous impact in the medical field. In particular, indocyanin green (ICG) has been widely used for assessing tumor margins during intraoperative procedures. Typically, the dye is intravenously injected into the patient, and after 24 hours the dye is removed from the patient’s body, except where binding between the dye and tumor cells has occurred. This selective binding between ICG and cancerous tissue allows for easy and accurate detection of cancer margins as well as detection of metastasis throughout the patient’s body. In order to detect the binding sites, a …

Identifying Humans By The Shape Of Their Heartbeats And Materials By Their X-Ray Scattering Profiles, Ikenna C. Odinaka

McKelvey School of Engineering Theses & Dissertations

Security needs at access control points presents itself in the form of human identification and/or material identification. The field of Biometrics deals with the problem of identifying individuals based on the signal measured from them. One approach to material identification involves matching their x-ray scattering profiles with a database of known materials.

Classical biometric traits such as fingerprints, facial images, speech, iris and retinal scans are plagued by potential circumvention they could be copied and later used by an impostor. To address this problem, other bodily traits such as the electrical signal acquired from the brain (electroencephalogram) or the heart …

Miniaturization Of Electrical Ultrafine Particle Sizers, Siqin He

McKelvey School of Engineering Theses & Dissertations

Nanoparticles, or ultrafine particles, have potential risks for human health, and the adverse health effects caused by ultrafine particles have been proven to be size related. To meet the increasing demanding for personal exposure monitoring and spatial distribution measurements of ultrafine particles, this dissertation studied the development and miniaturization of electrical ultrafine particle sizers (EUPS). There are three essential components for developing a EUPS unit: a charger to electrically charge the sample particles, an electrical mobility classifier to classify the charged particles, and a downstream particle count detector to measure the number concentrations. Two generations of EUPS were developed in …

Global Edf Scheduling For Parallel Real-Time Tasks, Jing Li

McKelvey School of Engineering Theses & Dissertations

As multicore processors become ever more prevalent, it is important for real-time programs to take advantage of intra-task parallelism in order to support computation-intensive applications with tight deadlines. In this thesis, we consider the Global Earliest Deadline First (GEDF) scheduling policy for task sets consisting of parallel tasks. Each task can be represented by a directed acyclic graph (DAG) where nodes represent computational work and edges represent dependences between nodes. In this model, we prove that GEDF provides a capacity augmentation bound of 4-2/m and a resource augmentation bound of 2-1/m. The capacity augmentation bound acts as a linear-time schedulability …

A Study Of Sampling Strategies For Helical Ct, Daheng Li

McKelvey School of Engineering Theses & Dissertations

Two classes of subsampling strategies, partially inspired by ideas from compressed sensing (CS), are developed and tested using real medical x-ray CT data acquired with a helical geometry. A version of the Feldkamp algorithm for helical x-ray CT is described. An alternating minimization (AM) algorithm for finding the maximum-likelihood estimates of attenuation functions in transmission X-ray tomography, developed by O’Sullivan and Benac, is then introduced. The derivation of this AM algorithm is extended to include an optional regularization term, which makes it a MAP estimate. A Newton’s method with trust region modification is implemented for the regularization. In addition, the …

Dynamic Thermal And Power Management: From Computers To Buildings, Yong Fu

McKelvey School of Engineering Theses & Dissertations

Thermal and power management have become increasingly important for both computing and physical systems. Computing systems from real-time embedded systems to data centers require effective thermal and power management to prevent overheating and save energy. In the mean time, as a major consumer of energy buildings face challenges to reduce the energy consumption for air conditioning while maintaining comfort of occupants. In this dissertation we investigate dynamic thermal and power management for computer systems and buildings. (1) We present thermal control under utilization bound (TCUB), a novel control-theoretic thermal management algorithm designed for single core real-time embedded systems. A salient …

Engineering Poly(Ethylene Glycol) Materials To Promote Cardiogenesis, Amanda Walker Smith

McKelvey School of Engineering Theses & Dissertations

Heart failure is one of the leading causes of death worldwide, and the current costs of treatment put a significant economic burden on our societies. After an infarction, fibrotic tissue begins to form as part of the heart failure cascade. Current options to slow this process include a wide range of pharmaceutical agents, and ultimately the patient may require a heart transplant. Innovative treatment approaches are needed to bring down costs and improve quality of life. The possibility of regenerating or replacing damaged tissue with healthy cardiomyocytes is generating considerable excitement, but there are still many obstacles to overcome. First, …

Improving Radiotherapy Targeting For Cancer Treatment Through Space And Time, Camille Noel

McKelvey School of Engineering Theses & Dissertations

Radiotherapy is a common medical treatment in which lethal doses of ionizing radiation are preferentially delivered to cancerous tumors. In external beam radiotherapy, radiation is delivered by a remote source which sits several feet from the patient's surface. Although great effort is taken in properly aligning the target to the path of the radiation beam, positional uncertainties and other errors can compromise targeting accuracy. Such errors can lead to a failure in treating the target, and inflict significant toxicity to healthy tissues which are inadvertently exposed high radiation doses.

Tracking the movement of targeted anatomy between and during treatment fractions …

Generating High Purity Embryonic Stem Cell-Derived Cell Populations For Transplantation Following Spinal Cord Injury, Dylan A. Mccreedy

McKelvey School of Engineering Theses & Dissertations

Embryonic stem cells hold great potential for cell replacement strategies in the central nervous system. Pre-differentiation into various neural cell types can help generate tissue-specific cell populations that can replace cells and tissue lost due to do injury or disease. A small number of undifferentiated pluripotent stem cells persist in most transplant populations even after pre-differentiation. Given the right environment, i.e. biomaterial scaffolds, these cells can lead to tumor formation thereby eliminating any potential therapeutic benefit. This dissertation focused on the development of high purity embryonic stem cell-derived cell populations devoid of pluripotent stem cells for transplantation into the central …

Polarization Imaging Sensors In Advanced Feature Cmos Technologies, Raphael Njoroge Njuguna

McKelvey School of Engineering Theses & Dissertations

The scaling of CMOS technology, as predicted by Moore's law, has allowed for realization of high resolution imaging sensors and for the emergence of multi-mega-pixel imagers. Designing imaging sensors in advanced feature technologies poses many challenges especially since transistor models do not accurately portray their performance in these technologies. Furthermore, transistors fabricated in advanced feature technologies operate in a non-conventional mode known as velocity saturation. Traditionally, analog designers have been discouraged from designing circuits in this mode of operation due to the low gain properties in single transistor amplifiers. Nevertheless, velocity saturation will become even more prominent mode of operation …

Calcium Cycling Disturbances And Arrhythmogenesis, Namit Gaur

McKelvey School of Engineering Theses & Dissertations

In this work, a detailed multiscale computational model of Ca cycling and action potential in a ventricular myocyte is developed and used to study mechanisms of arrhythmias associated with mutations in the ryanodine receptor (RyR). The multiscale ventricular myocyte model reproduces experimentally observed Ca dynamics both at the local dyadic scale and at the global whole-cell scale, while also simulating the action potential (AP) shaped by membrane ionic currents. The model represents stochastic activation of L-type Ca channels (LCCs) and RyRs in the dyads and simulates random generation of Ca sparks both during excitation-contraction coupling (ECC) and during diastole. The …

Atomic Transfer For Distributed Systems, Haraldur Darri Thorvaldsson

McKelvey School of Engineering Theses & Dissertations

Building applications and information systems increasingly means dealing with concurrency and faults stemming from distribution of system components. Atomic transactions are a well-known method for transferring the responsibility for handling concurrency and faults from developers to the software's execution environment, but incur considerable execution overhead. This dissertation investigates methods that shift some of the burden of concurrency control into the network layer, to reduce response times and increase throughput. It anticipates future programmable network devices, enabling customized high-performance network protocols.

We propose Atomic Transfer (AT), a distributed algorithm to prevent race conditions due to messages crossing on a path of …