Treasures @ UT Dallas
Not a member yet
7697 research outputs found
Sort by
An Exploration of Computational Design in the Arts
No full textThis Master of Fine Arts thesis paper, An Exploration of Computational Design in the Arts,
investigates how computational design tools are being integrated in modern artistic practice. The
artwork of the author, Travis Fowler, and adjacent contemporary artists will be investigated. The
method of evaluation utilizes a three-pronged approach: material, process, meaning. This form of
critique was developed in the 3D Studio at The University of Texas at Dallas and will be the
foundation for investigating each project described in this paper. Finally, this document
highlights the educational and professional journey of the author, his process of refining a
creative practice and select projects from his pursuit of a Master of Fine Arts degree
Characterizing the Link Between Behavior and Neurobiology of Verbal Working Memory in Young Adults With and Without Developmental Language Disorder
No full textDeficits in verbal working memory commonly occur in developmental language disorder (DLD)
and are hypothesized to contribute to the language impairment due to its importance in early
childhood language acquisition (Ellis Weismer et al., 1999; Evans et al., 2018; Gathercole, 1998).
Individuals with DLD perform poorly on verbal working memory tasks (e.g., nonword repetition,
n-back, sentence- and word-span tasks), and this performance correlates to linguistic proficiency
in lexical and sentence-level comprehension. While deficits in verbal working memory and
language are well documented in DLD, due to the current diagnostic methods, it has been debated
whether individuals with DLD represent a distinct clinical population, as opposed to the bottom
percentage of language users in the normal population (Dollaghan, 2004, 2011). Recently, research
has shown that a mismatch between neural activation patterns and behavior may be a potential
neurobiological marker of DLD (Brown et al., 2014; Haebig et al., 2017). To examine whether
DLD represents a true clinical population or is rather representative of low language ability typical
individuals, this manuscript-based dissertation contains three studies examining the pattern of
vii
cortical activation across the prefrontal cortex (PFC) for verbal working memory in neurotypical
adults and individuals with developmental language disorder (DLD) using functional near infrared
spectroscopy (fNIRS). The first study (Chapter 3) identified the neural response across the PFC
during an auditory 2-back task in adults with no history of language or learning disorders. Results
from this study showed differences in the change in hemoglobin level patterns for normal language
adults in response to high- and low-frequency words, as well as faster reaction times for low-, and
greater accuracy for high-frequency words. The second study (Chapter 4) expanded upon the
findings of the first study and showed that while two adults with DLD had behavioral performance
within normal limits of a group of age- and gender-matched controls, individual z-score analysis
revealed the adults with DLD exhibited qualitative differences in HbO2 and Hb in response to
high- and low-frequency words in the PFC as compared to normal controls. In addition, the second
study compared the adults with DLD to a low-performing normal language control, showing that
while both DLD subjects had increased levels of hemoglobin across the PFC, the low-performing
normal language control displayed significant reduction in hemoglobin levels across the PFC in
response to word frequency in the verbal working memory task. The findings of this study suggest
that not only do the individuals with DLD have atypical neural response to word frequency during
a verbal working memory task that distinguishes them from normal language peers, but also that
their atypical neural response does not resemble that of the low-performing normal language
population. The third study (Chapter 5) examined the ability of fNIRS to monitor changes in
hemoglobin levels across the PFC during the auditory 2-back task both before and after a
stimulation-based training that combined high-definition transcranial direct current stimulation
(HD tDCS) with a phonological working memory task. Using a single-subject design this study
confirmed the feasibility of fNIRS to track individual differences in changes in brain activity
immediately before and after working memory training and showed that neural response for both
individuals with DLD moved to within the expected range similar to that of the normal controls
after training.
Collectively these three studies enhance our understanding of the relationship between atypical
neural activation patterns paired with normal behavior in adults with DLD to address the
theoretical question of whether DLD represents a true clinical population, or merely low-normal
language users. The findings support a clinical definition of DLD and offer a potential neural
biomarker to detect these individuals that may otherwise go undetected due to their high functional,
but compensatory, performance
SPACs and Risk Factor Disclosures
No full textDuring 2020 and 2021, Special Purpose Acquisition Companies (“SPACs”) became a popular
alternative to going public, especially for young and risky firms. This study investigates the
informativeness of risk factor disclosures in SPACs’ proxy statements. I find that SPACs with
more specific risk information disclosed in the proxy statements have greater redemption rates.
This result suggests that shareholders are more likely to withdraw their capital in response to more
specific risk factor disclosures. The cross-sectional analyses show that the positive association
between the specific risk factor disclosures and redemption rates exists particularly in SPACs with
sponsor teams having less PE/VC or CEO experience, and SPACs with more retail investor
ownership. I also document that SPACs with more specific risk factor disclosures have greater
post-merger stock return volatility but are less likely to experience a significant decrease in net
income during post-merger periods. Overall, the findings suggest that SPAC shareholders
incorporate specific risk factor disclosures for their decision-making; however, they may overreact
to these specific risk factor disclosures and interpret such information incorrectly
Set-based Hierarchical Control for Multi-timescale Energy Management
No full textThe high-level goal of this PhD dissertation is to develop novel controller formulations and
analysis techniques to provide provable closed-loop system behaviors for complex systems.
Specifically, this dissertation focuses on the use of novel set-based mechanisms in the development of hierarchical Model Predictive Control (MPC) formulations for multi-timescale
energy management systems. Many complex systems such as hybrid-electric vehicle energy
systems, smart power grids, water distribution networks, and Heating Ventilation and Air-
Conditioned (HVAC) systems are multi-timescaled and have long operation times. A single
centralized MPC controller with fast update rates and a long prediction horizon might not be
able to solve the control optimization problem within the allocated time and thus, real-time
control actuation is not possible.
Alternatively, a hierarchical control architecture can be used to distribute control decisions
among various controllers connected in a hierarchy where the upper-level controller plans
coarse state and input trajectories at slow time steps while the lower-level controllers utilizes
this information to plan state and input trajectories at fast time steps. However, existing hierarchical formulations are not well suited to maximize system transient performance subject
to state, input, and terminal constraints. The open research challenge is how to provide the
flexibility to the lower-level controllers through novel coordination mechanisms to maximize
control performance while guaranteeing state, input, and terminal constraint satisfaction.
To achieve a computationally efficient hierarchical MPC algorithm for multi-timescale energy
management, the following research problems are addressed in this dissertation.
1. Development of a multi-level vertical hierarchical MPC framework for systems with additive known and unknown bounded disturbances. The proposed hierarchical control
algorithm is proven to be recursively feasible and is scalable with increase in prediction
horizon and number of states. The sub-optimality index of the hierarchical controller
is enhanced through a wayset and terminal cost-based coordination. To facilitate a
computationally efficient hierarchical control algorithm, set computations have been
developed for zonotopes and constrained zonotopes with a focus on application to systems and control.
2. Development of a tube-based robust MPC with simultaneous optimization of uncertainty sets using zonotopes. The proposed control formulation guarantees recursive
feasibility and constraint satisfaction to bounded additive disturbances from an uncertainty set optimized in real-time. The control formulation is extended to a full
hierarchical MPC where the uncertainty is quantified based on difference in control
decisions between hierarchical levels and between controllers in the same level. The
hierarchical control framework is shown to be recursively feasible and guarantees state
and input constraint satisfaction.
3. Development of a wayset-based Stochastic MPC framework that guarantees Mission-
Wide Probability of Safety (MWPS) for systems with long duration. To enable longer
missions under greater uncertainty, the wayset-based stochastic MPC allows for the
prediction horizon of the MPC to be significantly shorter than the length of the mission.
A scenario-based approach is used to approximate the stochastic MPC formulation and
recursive feasibility is proven
Graph Neural Networks for Property-guided Molecule Generation
No full textOrganic Photovoltaic Molecules (OPVs) have attracted chemists’ attention in improving
their electricity production efficiency while maintaining a low production cost. Designing
more efficient OPVs is slow and challenging due to the larger and more complex structures
of OPVs compared to other molecules. Moreover, there are currently no large-scale datasets
of inefficient and efficient versions of OPVs for the supervised generation of more efficient
molecules. Hence, we formulate this molecule design task as an unsupervised, propertyguided molecule optimization task, such that the chemical properties of the generated OPV
candidate molecules closely match the desired property values for high efficiency. Specifically,
we propose a motif-based graph-to-graph generative model to address the large molecule size
of OPVs. The generated OPV candidates are, according to chemists, rated promising with
minimal modifications
Dimensions of Cybersecurity: Espionage, Rivalry, and Political Economy
No full textThis dissertation consists of three interrelated projects and explores how cybersecurity affects
trade relations, espionage patterns, and militarized conflict behavior. My first project explores
data localization measures as a response to shocks of severe cyber incidents. I argue that severe
cyber incidents are punctuating events that shock the system, bringing the issue of strengthened
data protection to the policymaking agenda. To test this argument, I created a novel database of
data localization measures for 163 states between 2000-2020. My second project focuses
specifically on China, and how it uses cyber espionage to change its geopolitical environment.
Using a text-as-data approach, I hand-coded data found in 600 websites, articles, and technical
reports to disaggregate Chinese espionage from broader data sets, allowing me to test hypotheses
related to international and domestic politics. My last project explores whether cyber operations
are used as complements or substitutes (or neither) to conventional militarized confrontations. I
argue that a unique logic exists for why a state would choose cyber over more traditional means
of conflict, and this depends, largely, on the rivalry status of a given pair of states. To test my
hypotheses, I examine cyber operations and militarized incidents between 2000-2014
Analysis of the Transcriptomics, Physiology, and Invasiveness of Uropathogenic and Non-pathogenic Escherichia Coli
No full textEscherichia coli is a disease-causing species that can be divided into several phylogenetic
groups (PGs). The B2 group of the Clermont classification scheme causes up to 65% of E.
coli caused urinary tract infections (UTIs), with the remaining 35% caused by the summation of
groups A, B1, and D. A common set of virulence genes has not been identified. To determine the
features of B2 strains that contribute to virulence, I analyzed 35 strains by comparative RNA
sequencing. The strains were non-pathogenic and pathogenic, i.e., isolated from individuals with
a UTI, from groups A, B2 and D. I established transcriptomic differences of core gene
expression between groups B2 and A/D that involve genes for all aspects of macromolecular
synthesis, pathways of energy generation, and environment-sensing transcription factors. I
propose that these differences are responsible for the B2 group’s enhanced virulence potential.
Virulence requires attachment to and the formation of intracellular bacterial colonies within
epithelial cells. I found that five of five uropathogenic and one of three nonpathogenic B2 strains
had 25-50 times more intracellular colonies than strains from either groups A or D. My analyses
of bacterial appendages and the receptors for bacteria on epithelial cells show multiple entry
mechanisms through which E. coli can utilize to invade human urothelium.
A UTI involves multiple environments where bacteria must grow, including urine. Analyzing
growth in urine can be difficult as urine is a highly unreproducible mixture, so pools of human
urine are commonly used to minimize individual variation. I analyzed how the bacteria respond
to growth in an individual’s urine and in pools of urine from multiple classes of patients. I
identified pooled urine as a highly variable environment that does not necessarily mimic the
average growth of urine from individual patients. These results suggest that better urine mimics
are needed for future studies and that variations in urinary nutrient content could affect the host-
pathogen interaction and the outcome of the infection. In the same study, I also observed that
group B2 uropathogens are better at obtaining low levels of nutrients.
Genetic analyses of uropathogens are difficult because many are resistant to a common form of
genetic exchange mediated by a virus, known as transduction. We developed a method by which
uropathogens will accept the transducing phage P1 and subsequently allow for alterations of the
pathogenic chromosome. With the use of this method, metabolic mutants were constructed, and
the use of the genes were studied in motility, another prominent virulence function
in uropathogens. This led to the identification of a mechanism of motility, using the pili
appendage, previously thought not to be used by E. coli. The regulation of this motility
mechanism and alanine synthesis were also studied in part to help identify novel treatment
avenues for UTIs.
In summary, I provided evidence that suggests the basis for group B2’s virulence: Group
B2 strains have an altered transcriptome, and increased invasiveness, metabolism, and nutrient
acquisition capabilities. A new approach to the transduction protocol allows genetic alterations
that allows easier genetic analysis of these prominent pathogens
Face Identity “Likeness”: Insights for the Study of Face Perception and Identification
No full textColloquially, we commonly observe that some face images look more “like” an identity than
others. This experience stems from the fact that different images of the same identity can
vary in appearance, and that this appearance variation affects how closely an image resembles our own internal representation of what that identity should look like. Although we
perceive the “likeness” of face identities on a daily basis, surprisingly little is known about
how these perceptions are formed. Do we perceive face images as a better likeness if they
are photographed a certain way? Are face images of an identity perceived as a better likeness if they resemble images of that identity which have been seen previously? Further, are
identities represented by prototypes that reflect the viewing experience an observer has with
that identity? In a set of experiments, I addressed each of these questions using a combination of psychological and computational methods. First, using face images of identities
that participants are unfamiliar with and wherein each identity is shown across the same
changes in viewpoint and illumination, I tested whether higher likeness ratings are assigned
to certain viewpoint or illumination conditions (Experiment 1). The results showed that
participants who are unfamiliar with a face identity rate images as a better likeness when
the images show the identity in a more frontal viewpoint and with flash (as opposed to
ambient) illumination. At profile viewpoints, there is no difference in the likeness ratings
assigned to face images across illumination conditions. Next, using an image-based “face
space” generated by processing face images through a deep convolutional neural network
trained for face identification, I tested whether participants assign higher likeness ratings to
face images that either a) resemble a “central identity prototype” of a given face identity,
or b) exist within a more dense region of that identity’s specific subspace within the overall
DCNN-generated face space (Simulation 1). This simulation demonstrated that measures of
local area density are consistent with how human observers rate the perceived likeness of a
face image. Further, the distance of an image from an identity-specific prototype showing
the same identity was not consistent with human ratings of perceived likeness. Finally, by
familiarizing participants with an identity using images that show the identity from a single viewpoint or illumination, I tested whether participants assign higher likeness ratings to
images that resemble those which were seen previously for a given identity (Experiments 2
and 3). The results showed that, regardless of either the viewpoint/illumination of the face
image being rated and the viewpoint/illumination of face images that were seen previously
of a given identity, participants rate images as a better likeness if the image they are rating
matches the viewpoint/illumination of the images they were shown previously. Collectively,
these experiments provide insight into how variation in appearance is perceived across images of a given identity, and how this variation contributes to a face image being perceived
as a good likeness of the identity being portrayed
Hyperline Operator
No full textA postmortem development diary/retrospective of my growth as a game designer, and overview
of my thesis project, Hyperline Operator
Toward Practical Automatic Program Repair
No full textAutomatic program repair (APR) is one of the recent advances in automated software engineering aiming for reducing the burden of debugging by suggesting patches that either
directly fix the bugs or help the programmers during manual debugging. Despite the remarkable progress of APR in the last decade, state-of-the-art techniques suffer from problems in
three areas of scalability (i.e., handling large systems), applicability (i.e., handling different
programming languages), and effective patch correctness assessment (i.e., combating weak
specification problem of test suites), reducing practicality of APR.
In this dissertation, we take steps toward realizing practical APR by proposing solutions
to alleviate each of the above-mentioned problems. As for scalability and applicability,
we introduce and evaluate Java Virtual Machine (JVM) bytecode-level patch generation
and validation which allows (1) on-the-fly patch generation and validation and (2) uniform
treatment of programs written in dozens of programming languages. Through on-the-fly patch
generation and validation, we bypass many expensive steps in transforming programs making
our technique 10+X faster than state-of-the-art. This speed-up, in turn, allows our technique
to explore more of repair search space and find more genuine fixes than state-of-the-art fixing
55 (out of 587) Defects4J bugs. We provide empirical evidence on the applicability of our
technique on programming languages other than Java by applying it on 118 Kotlin bugs
from Defexts data set, out of which 14 was fixed.
We also introduce and evaluate a technique for correctness assessment of automatically
generated patches via both ranking and classification. Our technique is based on the idea
that the buggy program is almost correct insofar as fixing bugs involves small changes to
the code and does not remove the code implementing correct functionality of the program.
Thus, we measure the impact of patches on both production code (via syntactic and semantic
similarity) and test code (via code coverage) to separate the patches that result in similar
programs and that do not delete desired program elements.
We evaluated our technique on 1,290 patches, generated by 29 Java-based APR systems for
Defects4J programs. The technique outperforms state-of-the-art raking and classification
techniques. Specifically, in 43% (66%) of the cases, it ranks the correct patch in top-1 (top2) positions, and in classification mode, the technique achieves an accuracy and F1-score of
0.855 and 0.846, respectively.
We implement all these ideas in an integrated framework named PRF which can both be
used as an APR tool and as a framework for developing novel research prototypes