Computer Sciences Commons^™

Novel Architectures And Optimization Algorithms For Training Neural Networks And Applications, Vasily I. Zadorozhnyy

Theses and Dissertations--Mathematics

The two main areas of Deep Learning are Unsupervised and Supervised Learning. Unsupervised Learning studies a class of data processing problems in which only descriptions of objects are known, without label information. Generative Adversarial Networks (GANs) have become among the most widely used unsupervised neural net models. GAN combines two neural nets, generative and discriminative, that work simultaneously. We introduce a new family of discriminator loss functions that adopts a weighted sum of real and fake parts, which we call adaptive weighted loss functions. Using the gradient information, we can adaptively choose weights to train a discriminator in the direction …

Smart Decision-Making Via Edge Intelligence For Smart Cities, Nathaniel Hudson

Theses and Dissertations--Computer Science

Smart cities are an ambitious vision for future urban environments. The ultimate aim of smart cities is to use modern technology to optimize city resources and operations while improving overall quality-of-life of its citizens. Realizing this ambitious vision will require embracing advancements in information communication technology, data analysis, and other technologies. Because smart cities naturally produce vast amounts of data, recent artificial intelligence (AI) techniques are of interest due to their ability to transform raw data into insightful knowledge to inform decisions (e.g., using live road traffic data to control traffic lights based on current traffic conditions). However, training and …

High Performance Data Acquisition And Analysis Routines For The Nab Experiment, David Mathews

Theses and Dissertations--Physics and Astronomy

Probes of the Standard Model of particle physics are pushing further and further into the so-called “precision frontier”. In order to reach the precision goals of these experiments, a combination of elegant experimental design and robust data acquisition and analysis is required. Two experiments that embody this philosophy are the Nab and Calcium-45 experiments. These experiments are probing the understanding of the weak interaction by examining the beta decay of the free neutron and Calcium-45 respectively. They both aim to measure correlation parameters in the neutron beta decay alphabet, a and b. The parameter a, the electron-neutrino correlation coefficient, is …

Awegnn: Auto-Parametrized Weighted Element-Specific Graph Neural Networks For Molecules., Timothy Szocinski, Duc Duy Nguyen, Guo-Wei Wei

Mathematics Faculty Publications

While automated feature extraction has had tremendous success in many deep learning algorithms for image analysis and natural language processing, it does not work well for data involving complex internal structures, such as molecules. Data representations via advanced mathematics, including algebraic topology, differential geometry, and graph theory, have demonstrated superiority in a variety of biomolecular applications, however, their performance is often dependent on manual parametrization. This work introduces the auto-parametrized weighted element-specific graph neural network, dubbed AweGNN, to overcome the obstacle of this tedious parametrization process while also being a suitable technique for automated feature extraction on these internally complex …

Predicting Material Properties: Applications Of Multi-Scale Multiphysics Numerical Modeling To Transport Problems In Biochemical Systems And Chemical Process Engineering, Tom Pace

Theses and Dissertations--Physics and Astronomy

Material properties are used in a wide variety of theoretical models of material behavior. Descriptive properties quantify the nature, structure, or composition of the material. Behavioral properties quantify the response of the material to an imposed condition. The central question of this work concerns the prediction of behavioral properties from previously determined descriptive properties through hierarchical multi-scale, multiphysics models implemented as numerical simulations. Applications covered focus on mass transport models, including sequential enzyme-catalyzed reactions in systems biology, and an industrial chemical process in a common reaction medium.

Maternal Proximity To Mountaintop Removal Mining And Birth Defects In Appalachian Kentucky, 1997-2003, Daniel B. Cooper

Theses and Dissertations--Public Health (M.P.H. & Dr.P.H.)

Background: Extraction of coal through mountaintop removal mining (MTR) alters many dimensions of the landscape, and explosive blasts, exposed rock, and coal washing have the potential to pollute air and water with substances known to increase risk of developmental and birth anomalies. Previous research suggests that infants born to mothers living in MTR coal mining counties have higher prevalence of most types of birth defects.

Objectives: This study seeks to examine further the relationship between MTR activity and birth defects by employing individual level exposure estimation through precise satellite data of MTR activity in the Appalachian region and maternal residence …

Computational Utilities For The Game Of Simplicial Nim, Nelson Penn

Theses and Dissertations--Computer Science

Simplicial nim games, a class of impartial games, have very interesting mathematical properties. Winning strategies on a simplicial nim game can be determined by the set of positions in the game whose Sprague-Grundy values are zero (also zero positions). In this work, I provide two major contributions to the study of simplicial nim games. First, I provide a modern and efficient implementation of the Sprague-Grundy function for an arbitrary simplicial complex, and discuss its performance and scope of viability. Secondly, I provide a method to find a simple mathematical expression to model that function if it exists. I show the …

Orthogonal Recurrent Neural Networks And Batch Normalization In Deep Neural Networks, Kyle Eric Helfrich

Theses and Dissertations--Mathematics

Despite the recent success of various machine learning techniques, there are still numerous obstacles that must be overcome. One obstacle is known as the vanishing/exploding gradient problem. This problem refers to gradients that either become zero or unbounded. This is a well known problem that commonly occurs in Recurrent Neural Networks (RNNs). In this work we describe how this problem can be mitigated, establish three different architectures that are designed to avoid this issue, and derive update schemes for each architecture. Another portion of this work focuses on the often used technique of batch normalization. Although found to be successful …

Unitary And Symmetric Structure In Deep Neural Networks, Kehelwala Dewage Gayan Maduranga

Theses and Dissertations--Mathematics

Recurrent neural networks (RNNs) have been successfully used on a wide range of sequential data problems. A well-known difficulty in using RNNs is the vanishing or exploding gradient problem. Recently, there have been several different RNN architectures that try to mitigate this issue by maintaining an orthogonal or unitary recurrent weight matrix. One such architecture is the scaled Cayley orthogonal recurrent neural network (scoRNN), which parameterizes the orthogonal recurrent weight matrix through a scaled Cayley transform. This parametrization contains a diagonal scaling matrix consisting of positive or negative one entries that can not be optimized by gradient descent. Thus the …

High-Order Integral Equation Methods For Quasi-Magnetostatic And Corrosion-Related Field Analysis With Maritime Applications, Robert Pfeiffer

Theses and Dissertations--Electrical and Computer Engineering

This dissertation presents techniques for high-order simulation of electromagnetic fields, particularly for problems involving ships with ferromagnetic hulls and active corrosion-protection systems.

A set of numerically constrained hexahedral basis functions for volume integral equation discretization is presented in a method-of-moments context. Test simulations demonstrate the accuracy achievable with these functions as well as the improvement brought about in system conditioning when compared to other basis sets.

A general method for converting between a locally-corrected Nyström discretization of an integral equation and a method-of-moments discretization is presented next. Several problems involving conducting and magnetic-conducting materials are solved to verify the accuracy …

Ten Simple Rules For Responsible Big Data Research, Matthew Zook, Solon Barocas, Danah Boyd, Kate Crawford, Emily Keller, Seeta Peña Gangadharan, Alyssa Goodman, Rachelle Hollander, Barbara A. Koenig, Jacob Metcalf, Arvind Narayanan, Alondra Nelson, Frank Pasquale

Geography Faculty Publications

No abstract provided.

Hybrid Parallelization Of The Nasa Gemini Electromagnetic Modeling Tool, Buxton L. Johnson Sr.

Theses and Dissertations--Electrical and Computer Engineering

Understanding, predicting, and controlling electromagnetic field interactions on and between complex RF platforms requires high fidelity computational electromagnetic (CEM) simulation. The primary CEM tool within NASA is GEMINI, an integral equation based method-of-moments (MoM) code for frequency domain electromagnetic modeling. However, GEMINI is currently limited in the size and complexity of problems that can be effectively handled. To extend GEMINI’S CEM capabilities beyond those currently available, primary research is devoted to integrating the MFDlib library developed at the University of Kentucky with GEMINI for efficient filling, factorization, and solution of large electromagnetic problems formulated using integral equation methods. A secondary …

A Physics-Based Approach To Modeling Wildland Fire Spread Through Porous Fuel Beds, Tingting Tang

Theses and Dissertations--Mechanical Engineering

Wildfires are becoming increasingly erratic nowadays at least in part because of climate change. CFD (computational fluid dynamics)-based models with the potential of simulating extreme behaviors are gaining increasing attention as a means to predict such behavior in order to aid firefighting efforts. This dissertation describes a wildfire model based on the current understanding of wildfire physics. The model includes physics of turbulence, inhomogeneous porous fuel beds, heat release, ignition, and firebrands. A discrete dynamical system for flow in porous media is derived and incorporated into the subgrid-scale model for synthetic-velocity large-eddy simulation (LES), and a general porosity-permeability model is …

An Optical Character Recognition Engine For Graphical Processing Units, Jeremy Reed

Theses and Dissertations--Computer Science

This dissertation investigates how to build an optical character recognition engine (OCR) for a graphical processing unit (GPU). I introduce basic concepts for both building an OCR engine and for programming on the GPU. I then describe the SegRec algorithm in detail and discuss my findings.

Modeling Of Spallation Phenomenon In An Arc-Jet Environment, Raghava Sai Chaitanya Davuluri

Theses and Dissertations--Mechanical Engineering

Space vehicles, while entering the planetary atmosphere, experience high loads of heat. Ablative materials are commonly used for a thermal protection system, which undergo mass removal mechanisms to counter the heat rates. Spallation is one of the ablative processes, which is characterized by the ejection of solid particles from the material into the flow. Numerical codes that are used in designing the heat shields ignore this phenomenon. Hence, to evaluate the effectiveness of spallation phenomenon, a numerical model is developed to compute the dynamics and chemistry of the particles. The code is one-way coupled to a CFD code that models …

Richardson Extrapolation-Based High Accuracy High Efficiency Computation For Partial Differential Equations, Ruxin Dai

Theses and Dissertations--Computer Science

In this dissertation, Richardson extrapolation and other computational techniques are used to develop a series of high accuracy high efficiency solution techniques for solving partial differential equations (PDEs).

A Richardson extrapolation-based sixth-order method with multiple coarse grid (MCG) updating strategy is developed for 2D and 3D steady-state equations on uniform grids. Richardson extrapolation is applied to explicitly obtain a sixth-order solution on the coarse grid from two fourth-order solutions with different related scale grids. The MCG updating strategy directly computes a sixth-order solution on the fine grid by using various combinations of multiple coarse grids. A multiscale multigrid (MSMG) method …

Data Assimilation And Visualization For Ensemble Wildland Fire Models, Soham Chakraborty

University of Kentucky Master's Theses

This thesis describes an observation function for a dynamic data driven application system designed to produce short range forecasts of the behavior of a wildland fire. The thesis presents an overview of the atmosphere-fire model, which models the complex interactions between the fire and the surrounding weather and the data assimilation module which is responsible for assimilating sensor information into the model. Observation plays an important role in data assimilation as it is used to estimate the model variables at the sensor locations. Also described is the implementation of a portable and user friendly visualization tool which displays the locations …