Treasures @ UT Dallas
Not a member yet
7697 research outputs found
Sort by
Abstractive Related Work Generation: a Span Based Approach
No full textThe literature review is a crucial part of conducting and publishing academic research. It
provides background of prior works and highlights the novelties of the current research. The
literature review is presented under different sections in different fields. In natural language
processing domain, it is present under the Related Work section. The task of automatically
generating a related work section consists of generating a series of citations given the rest
of the current research paper and a list of cited papers. Prior work focuses on generating
individual sentences for each cited paper. However, citations consist of variable-length text
units, which can be multiple sentences or just part of a sentence and can summarize more
than one cited paper. To address these issues, we build a dataset to extract citation spans,
text fragments influenced by cited papers, and distinguish the spans that describe the cited
paper in detail from those that only provide high-level concepts. We train a baseline citation generation model which generates masked spans given the context paragraph from the
related work section.
We also discuss the role of citation span length and propose a length-controlled summarization model that generates summaries of a user-specified length. We additionally propose
an approach where the generation length is predicted and used as the desired length by a
single model and discuss heuristic estimates of the desired length and how they impact the
performance of the generation model.
Finally, we discuss the importance of generating spans that rightly fit in the related work
context. Traditional abstractive summarization models often generate generic outputs that
do not fit well in the context. To force the model to pay attention to contextual information,
we propose a model which generates the context along with the masked span. We conduct a
case study and human evaluation to show how the proposed model improves the coherence
of the generated span
Three Chapters on Water Resource Management in Texas
No full textThese three chapters on water resource management in Texas highlight how water consumption may change in response to the various management and conservation efforts. These
observations of water consumption across different types of water use in multiple regions of
Texas help further contribute to the current literature on water conservation efforts due to
the uncertainties in water supply and growing demand. The analyses in my dissertation is
relevant to Texas water providers and others trying to tackle potential water shortages. In
my dissertation, I analyze how water management efforts differ for groundwater and shed
light on how water consumption and pricing may vary depending on different management
efforts. My dissertation starts with a comprehensive analysis of pricing practices in 423 municipalities across Texas from 2014 to 2020 and their impact on residential water consumption.
The dissertation also considers other demographic and climatic factors combined with the
pricing analysis that might reflect both the demand and supply side. Besides residential
water consumption, my dissertation also sheds light on water conservation efforts tailored
to irrigation water and further investigates their respective economic impacts. Moreover, it
is essential to note that water conservation efforts may also need to be tailored to different
water sources. My results together help provide water providers with more detailed factors
to consider when managing water differently across different uses and sources of water
Multi-antenna Millimeter-wave Radars: Algorithms and Performance Analysis
No full textMillimeter-wave (mmWave) radars with multi-antenna systems have become popular in numerous automotive and industrial sensing applications. For these applications, target estimation is a crucial function. However, accurately estimating a target’s parameters becomes
challenging due to either the limitations of the system parameters or the presence of clutter
and interference. To address these challenges, this dissertation focuses on developing robust
signal processing algorithms and studying their performance analyses.
Direction of arrival (DOA) estimation of a target is a fundamental problem for radar sensors.
Super-resolution algorithms like MUltiple SIgnal Classification (MUSIC) have been proposed
for better DOA estimation performance compared to classical approaches. MUSIC relies on
accurate partitioning of the eigenvectors of the spatial correlation matrix between the signal
eigenvectors (i.e., signal subspace) and noise eigenvectors (i.e., noise subspace). In the first
part of this dissertation, we present a novel statistical framework for analyzing the resolution
performance of the MUSIC algorithm in resolving two closely spaced targets according to
the number of noise eigenvectors used in the spectrum evaluation. Using this framework, we
derive an analytical expression for the probability of resolution of the MUSIC algorithm.
Multiple-input multiple-output (MIMO) radar achieves high angular resolution at the expense of certain limitations of its systems realized using different multiplexing techniques
such as time division multiplexing (TDM), frequency division multiplexing (FDM), and code
division multiplexing (CDM). Of all these multiplexing techniques, TDM is the appropriate
choice for automotive applications due to its low hardware complexity. The latter part of
this dissertation focuses on the Doppler ambiguity problem associated with a standard TDM
MIMO radar. In a standard TDM MIMO radar, transmitters are activated sequentially according to their natural spatial order. The drawback of the standard TDM MIMO approach
is the coupling of velocity and DOA information of the targets. The coupling reduces the
unambiguous estimation interval of the Doppler frequencies of the targets by the number
of transmit antennas being multiplexed. To solve this problem, we propose a novel cost
function for jointly estimating the Doppler frequency and DOA of the targets.
In the last part of this dissertation, we address the mutual interference problem between
automotive radars. With the increasing demand for radar sensors in automotive applications,
this mutual interference between them is inevitable due to their unregulated transmissions.
To reliably estimate the target’s parameters, this interference needs to be detected and
mitigated. To mitigate automotive interference, we propose a two-stage signal decomposition
approach
Electromagnetic Optimization of Switched Reluctance Motor for Torque Ripple and Vibration Mitigation
No full textSwitched reluctance motor (SRM) generates torque based on the principle of reluctance torque
using a discontinuous rotating magnetic field. Double saliency of SRM causes magnetic
reluctance to change with respect to rotor position. SRM is singly excited on the stator and it
does not need magnetic excitation on its rotor. This feature makes SRM to be a simple, low
cost, and robust configuration that makes it desirable for high speed and harsh applications.
However, SRM exhibits high levels of torque ripple contributing to its acoustic response. The
main contributing factor to this behavior is the non-uniform distribution of the flux and force
density in SRM. To elaborate, SRM experiences a sudden rise in the flux density, in the airgap
when rotor and stator poles start to overlap. This causes a sudden rise in the force density in
both tangential and radial components of force at points close to the stator slot and that leads
to the vibration and torque ripple.
To address this problem, a novel rotor geometry with optimally designed flux barriers has
been proposed in this dissertation to be used along with a conventional SRM stator. An
optimization algorithm comprised of Genetic Algorithm (GA) and Finite Element Analysis
(FEA) has been used to identify the best rotor geometry for maintaining average torque while
minimizing torque ripple and tangential vibration of the stator. The performance of the
optimized motor is then compared with a conventional SRM of the same size through
experiments. The results show significant improvement in torque ripple as well as vibration
for the new topology with no tangible drop in efficiency at high speeds
Supercapacitor Electrode Materials Comprising Uniformly Dispersed Chromium Nitride/ Carbon Fiber Composite
No full textNowadays, researchers and industrial designers are looking for an eco-friendly alternative energy
source to fulfill the increasing need for energy and reduce environmental pollution. Electricity
based on energy storage devices can be a way of solving the crisis. Among different energy storage
devices, the faster charging and discharging speeds or higher power densities, working in a wider
range of temperatures, and longer cycle life of supercapacitor make them attractive for several
applications. The commercially available supercapacitors are mostly electric double-layer
capacitors (EDLCs) and to a lesser degree of pseudocapacitors. Recently, researchers have been
focusing on hybrid supercapacitors (HSCs) due to their ability to combine the properties of both
EDLCs and pseudocapacitors to expand the applications. Engineered carbon nanofibers can be
coupled with conductive metal nitrides to form composites and used as electrode materials for
supercapacitor applications. In this work, a new hybrid nanocomposite of carbon fibers and
chromium nitride (CFs/CrN) was fabricated as electrode materials, where polyacrylonitrile (PAN)
was utilized as the carbonizing materials and polymethyl methacrylic acid (PMAA) as the
sacrificial agents. Here, pore-forming agents, PMAA, assisted in improving the supercapacitor's performance by increasing the electrode materials' surface area. Also, growing CrN nanoparticles
in the fiber contributed by the pseudocapacitance from proton adsorption. Furthermore, the
chelation ability of the PMAA might be beneficial for the homogeneous distribution of CrN all
over the CFs. Furthermore, the use of aqueous electrolytes and comparatively low-cost transition
metal materials lowered the fabrication costs, and the utilization of the electrospinning technique
makes CFs/metal nitrides composite electrodes freestanding and readily produced.
Chapter 1 describes a detailed introduction to supercapacitors, including a brief description of the
storage principle of EDLCs, pseudo capacitors, and hybrid supercapacitors and their advantages
and disadvantages. It also describes the basics of the electrospinning process, thermal treatments,
and aqueous electrolytes, all of which were applied to fabricate the supercapacitors using CFs and
metal nitrides composite-based electrodes.
Chapter 2 represents the fabrication of CFs and chromium nitrides composites using polymer
blends containing PAN and PMAA and Cr precursor as a source of chromium. This chapter also
describes the characterization of synthesized PAM-PMAA-CrN electrode materials and their
electrochemical performance and analysis. The highest capacitance was obtained from PANPMAA-CrN based electrode 159 F/g at 5 mV/s. Also, the highest energy densities of 13.26 Wh/Kg
at 1.2 V were obtained from the PAN-PMAA-CrN and CNFs based asymmetric device.
Furthermore, the PAN-PMAA-CrN electrode showed higher stability with 80.4% capacitance
retention after 10000 cycles
Characterization and Circuit Design of Soft Bend Sensors for Use in Robotic Hands and Orthotics
No full textSmart materials such as shape memory alloy (SMA) and twisted and coiled polymer fishing line
(TCPFL) are essential elements for the realization of novel smart robotic hands, orthotic hands,
and prosthetic hands. These artificial muscle-actuated robotic hands need to be assessed
extensively to understand the properties and efficiency of the designs. Cyclic movement of the
fingers must be monitored to characterize the actuation frequency of the artificial muscles and
the response due to the amplitude of stimuli. Flex sensors and strain gauges are commonly used
to observe the bending action of robotic fingers by attaching them along the finger’s length. Such
sensors are piezoresistive in nature and change their resistance due to stresses exerted on them
which change the sensor’s dimensions. This piezoresistive property of some standard sensors
available in the market is studied in this research to determine the angular position of the robotic
finger during flexion or extension when the artificial muscles are triggered. A similar type of
strain gauge sensor was designed in-house, which can be 3D printed and directly embedded into
an orthotic finger. This sensor was fabricated using conductive and soft filament materials which
consist of a composite of thermoplastic polyurethane (TPU) and carbon nanotubes (CNT). The
purpose of this thesis is to characterize the sensor and design the optimized signal conditioning
circuit of this sensor. A voltage divider circuit was utilized to characterize and understand the
properties of the sensors. Various techniques including Wheatstone bridge, differential
amplifiers, and active low pass filters have been implemented for designing the signal
conditioning circuit of the strain gauge sensor, and several simulations results were obtained.
This optimization converts the variable resistance of this strain gauge into a linearized voltage
signal that is easier to monitor through common interfaces. In general, soft smart materials,
actuators, and sensors can transform the existing robotic hands to achieve more elegant,
lightweight, and modular designs
Efficient Design and Optimization of Artificial Neural Networks: SW and HW
No full textArtificial Neural Networks (ANNs) have achieved significant advancements in machine intelligence due to their ability to learn complex tasks. However, their deployment on devices with limited resources or real-time applications requires consideration of the network's physical size or area. High-Level Synthesis (HLS) is a design methodology that simplifies the implementation of complex algorithms like ANN design by using high-level programming languages. This work investigates the construction of efficient ANNs by combining techniques such as architecture search, pruning, quantization, and compression to optimize the network's architecture, input/output data size, and computation precision. The optimization framework minimizes the area of the equivalent SystemC ANN model by reducing the bit width allocated for either the neurons or weights. The primary focus is on the application of handwriting recognition using the MNIST dataset, which serves as a prototype problem for understanding neural networks in general. The contributions of this thesis include an explorer to perform architecture search and an optimization framework to minimize the network's area, providing valuable insights into accuracy and hardware efficiency trade-offs
Multimodal Medical Image Analysis Using Machine Learning
No full textWith extensive collections of data and evolved medical diagnostic imaging, including digitized
histopathology images, computer-aided detection for medical assessment has become feasible.
Clinicians and medical professionals can use automated computational models to detect
regions of interest and aid in diagnosis. They can be used to provide a second opinion at times
of uncertainty or used independently for reducing the load of the medical healthcare provider
on difficult and time-consuming tasks. The research presented in this dissertation focuses
on developing automated systems comprising detection, classification, survival prediction,
segmentation, and quantification tasks using machine learning and deep learning algorithms
for three medical problems. We use multiview images, images and clinical information, and
multisite images to solve these problems, overcoming the underlying challenges, including
limited data and lack of region annotations.
In the first problem, we develop a solution to assess craniosynostosis, a skull deformity, automatically. An automatic craniosynostosis detector can diagnose the malformation early,
particularly helping care providers with limited craniofacial expertise. We analyze 2D multiview images of healthy controls and infants with craniosynostosis to identify the disease using
computer-based classifiers. First, we develop a traditional machine learning (ML) with feature extraction multiview image-based classifiers, and next, we build a Convolutional Neural
Networks (CNNs) based classifier. The ML model has an accuracy of 91.7%, and the CNN
model has an accuracy of 90.6%. ML model performs slightly better than the CNN model,
probably due to the supremacy of the designed ML features in the craniosynostosis subtypes
differentiation and small image dataset availability for model development.
In the second problem, we classify a common type of cancer occurring in the soft tissue of
children named Rhabdomyosarcoma (RMS) as the correct subtype. The subtypes respond
to different treatments. Due to slight differences in the appearance of histopathology images,
manual classification is tedious and needs high expertise. We present a machine learningbased pipeline to automatically classify Rhabdomyosarcoma into three significant subtypes
using whole slide images (WSI). We train the model based on the knowledge of the class
associated with the WSIs. There are no manual annotations used for the model development.
Meanwhile, most related approaches to classify tumor needs manual regional or nuclear
annotation on WSI. We first divide a WSI into tiles and predict the class of each tile. Then
we convert tile-level predictions to WSI-level predictions using threshold and soft voting.
We achieved 94.87% WSI tumor subtype classification accuracy on a large and diverse test
dataset. Unlike related work, we achieved such accurate classification at 5X magnification
of WSI, using 20X or 10X for best results. The benefit of our approach is that training and
testing are performed computationally faster due to the lower image resolution.
Next, we solve the survival prediction by developing a novel survival predictor. Our proposed
method comprises two steps. First is the extraction of a whole slide feature map (WSFM),
and next, it is used to build the survival predictor. We divide a WSI into small tile images,
then extract the features for the tile image using InceptionV3 pretrained model. Next, we
reduce the dimension of the features by applying Principal Component Analysis (PCA) and
obtain a low dimension feature representation for the tiles. We store the tile features as
channel information and replace all the tiles with their PCA extracted features in place to
form WSFMs. The WSFM has the information of the entire tissue in the WSI and also
preserves the adjacency information of the tiles. Using the WSFMs as input we then build
Siamese survival convolutional neural network (SSCNN) which overcomes the small dataset
size problem pertinent in the existing methods. The SSCNN uses the multivariate clinical
features aling with the WSFM for predicting a survival score. We propose a novel modified
pairwise ranking loss with a bounded inverse term to train the SSCNN. The proposed method
does not need pixel-level annotations which is a notorious bottleneck for such studies and can
be easily adapted for any tumor being agnostic to the other model development parameters
like number of clusters. Experimental results in two different tumors, RMS and Glioblastoma
multiforme (GBM) brain cancer, validate the success of the proposed SSCNN compared to
other state-of-the-art survival predictors.
At last, we established a deep learning model (DLM) pipeline to assess tumor viability using
WSIs of the primary tumors and corresponding lung sections for 130 mice having breast
cancer. We developed an InceptionResNetV3 convolutional neural network (CNN) model
for detecting the viable and necrotic tumors and normal mammary tissue in the primary tumor WSIs. Then, we trained another CNN model by fine-tuning the first model to identify
the metastatic tumor and normal tissue in the lung sections. We created the tumor viability heatmap for each WSI using the predictions from the respective model and quantified
the tumor viability in each WSI. We measured the intraclass correlation between the manual tumor viability quantification and the DLM and obtained more than 0.97 correlation.
By providing the clinically relevant outcome parameter of tumor viability, this novel DLM
promises to become a standard tool in animal tumor models’ assessment
Citizen Scientish Hypothetical Astronomy
No full textThis MFA Thesis describes the inspiration and artistic process through which the work exhibited
in Citizen Scientish was made and how that process informs my assertion of the actual existence,
rather than mere visual metaphor, of natural celestial bodies and configurations of those bodies
depicted in my visual work, itself informed by observation of and meditation on anthropogenic,
terrestrial surfaces. The series of images created for this thesis establish a creative dialectical
discourse with the previously scientifically discredited Celestographic work of 19th-century
playwright, painter, and amateur astronomer August Strindberg (1849 – 1912) and with current
popular, though widely debated, theories of multiple universes.
This reassessment establishes that the images presented in my work, while done so in the context
of fine art, are representations of fact tantamount to astronomical and cosmological discovery
Student Haiku Competition entries for 2023
No full textEntries Submitted for The Eugene McDermott Library 2023 Annual Haiku ContestEntries for the Eugene McDermott Library 2023 Annual Haiku CompetitionEugene McDermott Librar