Physical Sciences and Mathematics Commons^™

Quantification Of Hydrologic Response To Forest Disturbance In Western U.S. Watersheds, Sara A. Goeking '92

Doctoral Dissertations

Forested watersheds produce more than half of the water supply in the United States. Forests affect how precipitation is partitioned into available water versus evapotranspiration. This dissertation investigated how water yield and snowpack responded to forest disturbance following recent disturbances in western U.S. forests during the period 2000-2019.

Chapter 2 systematically reviewed 78 recent studies that examined how water yield or snowpack changed after forest disturbances. Water yield and snowpack often increased after disturbance, but decreased in some circumstances. Decreased water yield was most likely to occur following disturbances that did not remove the entire forest canopy. It was also …

Non-Academic Support Math Faculty Members Provide In Developmental Accelerated And Corequisite Support Courses In California Community Colleges, David Vakil '92

Doctoral Dissertations

To guide practitioners of rapidly evolving developmental math reform in community colleges, this study surveyed California community college math faculty who taught accelerated developmental courses or corequisite support courses. The survey was conducted during the early implementation phase of both course types, during spring and fall 2018 terms. This study measured faculty’s self-reported provision of forms of non-academic support, frequency of implementation, and reasons faculty believed the support would help students succeed. The literature review guided grouping non-academic support into five forms: nurturing, helping students’ motivation, providing a growth mindset theory of intelligence, helping provide social integration, and helping to …

Finding Critical And Gradient-Flat Points Of Deep Neural Network Loss Functions, Charles Gearhart Frye '09

Doctoral Dissertations

Despite the fact that the loss functions of deep neural networks are highly non-convex, gradient-based optimization algorithms converge to approximately the same performance from many random initial points. This makes neural networks easy to train, which, combined with their high representational capacity and implicit and explicit regularization strategies, leads to machine-learned algorithms of high quality with reasonable computational cost in a wide variety of domains.

One thread of work has focused on explaining this phenomenon by numerically characterizing the local curvature at critical points of the loss function, where gradients are zero. Such studies have reported that the loss functions …

Acoustic Confinement And Characterization Of A Microwave Plasma, Seth Lee Pree '09

Doctoral Dissertations

High amplitude acoustic fields are used to confine, characterize, and manipulate collisional plasmas with temperatures of a few thousand Kelvin. This dissertation describes the theory, experimental techniques, and apparatus necessary both to generate high amplitude sound in a few thousand Kelvin plasma and to use that sound field to manipulate the plasma within a resonant acoustic cavity. The acoustic field in a spherically symmetric oscillating plasma has been measured to have a Mach number of .03, which is sufficient to cause acoustic radiation pressure effects to confine the plasma to the center of its container. This field also generates convection …

Equidimensional Adic Eigenvarieties For Groups With Discrete Series, Daniel Robert Gulotta '03

Doctoral Dissertations

We extend Urban's construction of eigenvarieties for reductive groups G such that G(R) has discrete series to include characteristic p points at the boundary of weight space. In order to perform this construction, we define a notion of "locally analytic" functions and distributions on a locally Q_p-analytic manifold taking values in a complete Tate Z_p-algebra in which p is not necessarily invertible. Our definition agrees with the definition of locally analytic distributions on p-adic Lie groups given by Johansson and Newton.

The Statistical Dynamics Of Nonequilibrium Control, Grant Murray Rotskoff '09

Doctoral Dissertations

Living systems, even at the scale of single molecules, are constantly adapting to changing environmental conditions. The physical response of a nanoscale system to external gradients or changing thermodynamic conditions can be chaotic, nonlinear, and hence difficult to control or predict. Nevertheless, biology has evolved systems that reliably carry out the cell’s vital functions efficiently enough to ensure survival. Moreover, the development of new experimental techniques to monitor and manipulate single biological molecules has provided a natural testbed for theoretical investigations of nonequilibrium dynamics. This work focuses on developing paradigms for both understanding the principles of nonequilibrium dynamics and also …

Transport Architectures For An Evolving Internet, Keith Winstein '99

Doctoral Dissertations

In the Internet architecture, transport protocols are the glue between an application’s needs and the network’s abilities. But as the Internet has evolved over the last 30 years, the implicit assumptions of these protocols have held less and less well. This can cause poor performance on newer networks—cellular networks, datacenters—and makes it challenging to roll out networking technologies that break markedly with the past.

Working with collaborators at MIT, I have built two systems that explore an objective-driven, computer-generated approach to protocol design. My thesis is that making protocols a function of stated assumptions and objectives can improve application performance …

Borel Complexity Of The Isomorphism Relation For O-Minimal Theories, Davender Singh Sahota '99

Doctoral Dissertations

In 1988, Mayer published a strong form of Vaught's Conjecture for o-minimal theories (1). She showed Vaught's Conjecture holds, and characterized the number of countable models of an o-minimal theory T if T has fewer than 2K ° countable models. Friedman and Stanley have shown in (2) that several elementary classes are Borel complete. This work addresses the class of countable models of an o-minimal theory T when T has 2N ° countable models, including conditions for when this class is Borel complete. The main result is as follows.

Theorem 1. Let T be an o-minimal theory in a countable …

Ads/Cft In String Theory And M-Theory, Daniel R. Gulotta '03

Doctoral Dissertations

The AdS/CFT correspondence is a powerful tool that can help shed light on the relationship between geometry and field theory.

The first part of this thesis will focus on the construction of theories dual to Type IIB string theory on AdS₅ x Y⁵, where Y⁵ is a toric Sasaki-Einstein manifold. This thesis will introduce a consistency condition called "proper ordering" and demonstrate that it is equivalent to several other previously known consistency conditions. It will then give an efficient algorithm that produces a consistent field theory for any toric Sasaki-Einstein Y⁵.

The second part …

Proton-Proton Correlation Functions As A Probe To Reaction, Micha A. Kilburn '95

Doctoral Dissertations

In an experiment at NSCL, proton-proton (p-p) correlation functions were measured in ⁴⁰Ca+⁴⁰Ca and ⁴⁸Ca+⁴⁸Ca reactions, both at E/A = 80 MeV. The High Resolution Array (HiRA) detected light particles with excellent energy (<200 keV) and angular (~0.2^o) resolution. The MSU 4pi Array covered 77% of the total 4pi solid angle and was used to determine the impact parameter for collisions using transverse energy (E_t) as the relevant observable.

Two-particle correlation functions are employed in this work to measure the space-time …

A Measurement Of The Top Quark Mass With A Matrix Element Method, Adam Paul Gibson '96

Doctoral Dissertations

We present a measurement of the mass of the top quark. Our event sample is selected from proton-antiproton collisions, at 1.96 TeV center-of-mass energy, observed with the CDF detector at Fermilab's Tevatron. We consider a 318 pb^-1 dataset collected between March 2002 and August 2004. We select events that contain one energetic lepton, large missing transverse energy, exactly four energetic jets, and at least one displaced vertex b tag. Our analysis uses leading-order tt and background matrix elements along with parameterized parton showering to construct event-by-event likelihoods as a function of top quark mass. From the 63 events observed …

Modular Machine Learning Methods For Computer-Aided Diagnosis Of Breast Cancer, Mia Kathleen Markey '94

Doctoral Dissertations

The purpose of this study was to improve breast cancer diagnosis by reducing the number of benign biopsies performed. To this end, we investigated modular and ensemble systems of machine learning methods for computer-aided diagnosis (CAD) of breast cancer. A modular system partitions the input space into smaller domains, each of which is handled by a local model. An ensemble system uses multiple models for the same cases and combines the models' predictions.

Five supervised machine learning techniques (LDA, SVM, BP-ANN, CBR, CART) were trained to predict the biopsy outcome from mammographic findings (BIRADS™) and patient age based on a …

Searches For Neutral Higgs Bosons In Quark-Antiquark Tau-Antitau Final States Using The Opal Detector At Lep, James Andrew Hocker '91

Doctoral Dissertations

Searches for neutral Higgs bosons have been performed with the OPAL detector at LEP. Approximately 170 /pb of e+e- collision data at a center-of-mass energy of 189 GeV have been used to search for the SM process e+e- to HZ as well as for the processes e+e- to hZ and e+e- to hA which occur in extended Higgs theories. The searches are sensitive to final states containing quarks and tau leptons, for which an artificial neural network for the identification of tau leptons was designed. The results have been combined with OPAL searches for other final states to obtain a …

The Dynamics Of Cyberspace: Examining And Modelling Online Social Structure, Brian S. Butler '89

Doctoral Dissertations

It has been proposed that online social structures represent new forms of organizing which are fundamentally different from traditional social structures. However, while there is a growing body of empirical research that considers behavioral aspects of online activity, research on online social structure structural remains largely anecdotal. This work consists of three papers that combine previous studies of traditional social structures, empirical analysis of longitudinal data from a sample of Internet listservs, and computational modeling to examine the dynamics of social structure development in networked environments.

The first paper (Title: When is a Group not a Group: An Empirical Examination …