Treasures @ UT Dallas
Not a member yet
7697 research outputs found
Sort by
A Combinatorial Viewpoint on Few-Shot Object Detection
No full textThis thesis explores submodular functions, which have been leveraged by the Machine Learn-
ing community in recent years, and proposes how to apply these functions to achieve per-
formance gains on the task of Few-Shot Object Detection. Broadly speaking, we focus on
three aspects:
1. Explore submodular functions and their properties.
2. Implement the State-Of-The-Art models for the Few-Shot Object Detection task.
3. Introduce a framework that leverages submodular functions to improve the perfor-
mance of any Few-Shot Object Detection model.
This thesis begins with an introduction to Few-Shot Object Detection and submodular func-
tions. We then introduce our framework of submodular combinatorial objectives for usage
in Few-Shot Object Detection. Next we set up the experiments for implementing standard
models and our framework. We review the empirical results obtained from our experiments
and finally conclude with key takeaways
Mapping Royal Matronage: Queen Sancia and Urban Topographies of Medieval Naples
No full textThis thesis investigates the five religious foundations variously affiliated with the Franciscan
Orders that Sancia of Majorca, Queen of Naples, established and fervently supported over the
course of three decades between the years 1312 and 1343. This constellation of foundations
comprises two double convents, Santo Corpo di Cristo (1312), (also known as Santa Chiara) and
Santa Croce e Trinità (1338), and three charitable institutions, Santissima Annunziata (1318),
Santa Maria Maddalena (1324), and Santa Maria Egiziaca (1342). This thesis explores Sancia’s
role as a woman patron of these foundations, investigating her agency in regard to their
conception, siting, and construction to their administration and financial endowment to their
functions as places of charitable work or dynastic memory. After a sustained study of each
complex, this thesis analyzes the foundations’ spatial positions within the medieval city itself and
in relation to each other and asks how Sancia’s patronage shaped Naples’s social and physical
topography. In turn, her religious foundations in Naples are set within broader patterns of royal
women’s patronage—or Matronage—in Sancia’s own natal and marital families and across
European courts. For the first time, this thesis puts the five religious institutions founded and
supported by Sancia in Naples in direct conversation with one another, reiterating the agency
Sancia wielded within her matronage. This thesis defines Matronage as women’s practices of
commissioning artworks or monuments within their socio-politically accepted gendered
construct and spheres of influence, oftentimes mirroring, adapting, and modeling the actions of
influential women within their lives, be them direct contemporaries, predecessors, or both
Bifurcations of Low-dimensional Systems With Discontinuities: a Case Study of Hysteresis in Systems With Lennard–jones Interaction Potential and Hopf Bifurcation in Lur’e System With Dry Friction
No full textThe fundamental mechanism of hysteresis in the quasistatic limit of multi-stable systems is
associated with transitions of the system from one local minimum of the potential energy to
another. In this scenario, as system parameters are (quasistatically) varied, the transition is
prompted when a saddle-node bifurcation eliminates the minimum where the system resides
in. The objective of the present work is to specify this generic mechanism for systems of
interacting particles assuming a natural single-well (Lennard-Jones) interaction potential for
each pair of particles. We show multi-stability and present details of hysteresis scenarios
with the associated bifurcations and transitions in a case study of constrained four-degrees-
of-freedom four particle systems on the plane. We also include the analysis of stability and
bifurcation analysis of piecewise smooth systems which process a connected continuum of
equilibrium states
Physical Sensing for Assessing Environmental Influences on Bird Distribution, Migration, and Demography: an Integrated Study of Avian Diversity, Vocalization Patterns, and Habitat Characteristics
No full textThis is an applied physics thesis that uses physical sensing to examine avian diversity, which
is an important component of global ecosystems. There is substantial evidence to indicate
that environmental changes profoundly impact the distribution, migration patterns, and
populations of birds. This research focuses on North American avian species, to investigate
the effects of climate change using data from the North American Breeding Bird Survey
(BBS) and the Christmas Bird Count (CBC). Additionally, a distributed physical sensing
system has been developed to explore the connections between avian diversity, vocal behav-
ior, and the ambient environment. The study also examines how the behavior of the dawn
chorus of different species of birds is influenced by physical environmental factors such as
meteorology, air quality, ambient illumination, and solar zenith angle.
By employing a multifaceted approach, this research seeks to offer valuable insights into
how avian populations adapt to changing climate conditions and the implications for their
ecological functions. The results of this study are of relevance for the conservation and
management of avian species in light of the rapidly evolving climate
Performance Evaluation of Various Electric Motor Drives for Electric Propulsion
No full textElectric vehicles (EVs) offer environmental benefits by reducing greenhouse gas emissions and
air pollution. Additionally, EVs contribute to energy efficiency and can lead to decreased
dependency on fossil fuels. Because of that, worldwide sales of EVs have increased exponentially
in the last decade. Therefore, choosing the right electric drive becomes crucially important. The
main objective of this dissertation is to compare various electric drives for electric propulsion in
terms of two main criteria, the first one is the performance comparison for base torque condition
at 12 A/mm2 current density and at 25A/mm2 current density for continuous and peak operating
conditions respectively and the second criteria is electromechanical efficiency and thermal
response of each machine at the same current density. The performance comparison has been done
in terms of average torque, torque and power densities, torque pulsation, weight, peak and running
efficiency, drive and traction efficiency, electromechanical efficiency, active cost, impact of high
speed in constant power region, and impact of operating Electric Drive under Electromagnetic
Pollution (EMI).
The drive and traction efficiencies were calculated using identical inverters in the control drives
for all machines under both operating conditions, employing Ansys Simplorer software. A detailed
thermal analysis was conducted with two different cooling arrangements: oil cooling in the air gap
and end winding cooling using cooling discs. Ansys Fluent software was utilized to evaluate the
temperature of the machines and impact of different environmental temperatures at both base and
peak operating conditions for all seven machines. Additionally, AVL Cruise M software was
utilized to compare the performance of these electric drives under the Worldwide Harmonized
Light-Duty Vehicles Test Cycle (WLTC) to simulate the urban, suburban, rural, and highway
driving conditions using the Nissan Leaf as the reference electric vehicle.
This study examined a range of conventional electric traction motor types, along with two
unconventional traction drives. The traction drives evaluated include the Interior Permanent
Magnet Synchronous Machine (IPMSM), Induction Machine (IM), PM-assist Synchronous
Reluctance Machine (PM-SyncRel), Synchronous Reluctance Machine (SyncRel), Switched
Reluctance Machine (SRM), Double Stator Switched Reluctance Machine (DSSRM), and Double
Stator PM Synchronous Machine (DS-PMSM). All the machines share the same outer stator
diameter and maintain an equal mass of magnets in the Permanent Magnet (PM) machines. This
study clearly demonstrates DSPM machine achieves high torque and power density, maintains
acceptable thermal limits, and completes the Drive cycle efficiently
‘Model Minority’ Writing: Asian America and the Literary Canon
No full text“Model minority” writing are texts that function to fortify the discourse of “ethnic writing” as different from “American writing.” The more clearly differentiated an ethnic writing is, the more it is perceived to be “authentic”—the underlying assumption is that the ethnic text and its producer, being authentically different, are objects of exoticization and are “always already known.” Being a criterion for ethnic literature, “authenticity” tends to canonize the works that exhibit strong identity politics that show their essential ethnic-ness. In short, the more defined the ethnic literature is from the canon of (white) American literature, the more valued it becomes to the diversified, multi-cultural, melting-pot literatures of the neoliberal capitalist society.
This dissertation employs the historical materialist approach to examine the changing condition for inclusion of Asian Americans from the time of Sui Sin Far and Otono Watanna up to the present moment with Viet Thanh Nguyen. In framing literary production as cultural labor, the writers need to be “flexible” to the changing demands of the ruling class. Jose Garcia Villa once shone as the posterchild of modernist poetry with Have Come, Am Here (1943), though his inflexibility to show his ethnic-ness buried him into forgottenness; the unionist Carlos Bulosan whose oppositional appeal in America is In the Heart (1946) that foregrounds the protagonist’s
socioeconomic disenfranchisement and Filipino-ness makes him newly acclaimed, ensuring his canonization into Asian American literature. When the demand for model minoritarianism shifted from assimilation to identitarianism, the American literary canon did not really become more inclusive as it seems—it just created sub-canons from which identities could be marketed as commodities in the neoliberal capitalist America
Equivariant Global Hopf Bifurcation in Abstract Nonlinear Parabolic Equations
No full textIn this Thesis, we study local and global symmetric Hopf bifurcation in abstract parabolic
systems by means of the twisted equivariant degree
Light Amplification Using Colloidal Quantum Shells Nanoparticles
No full textColloidal semiconductor nanocrystals (NCs), also known as quantum dots (QDs), have been
widely researched for their optical properties, such as size-dependent bandgaps, narrow band
emission, high luminescence, and nonlinearities. However, when electron-hole pairs (excitons)
are confined within the small volume of the dot upon absorption, luminescence is significantly
compromised due to nonradiative Auger recombination. In this study, we introduce novel
colloidal quantum shells (QSs) with an “inverted” QD geometry, which effectively suppress
Auger decay. Transient Absorption (TA) analysis demonstrates the successful suppression
of biexciton (BX) and multiexciton (MX) Auger recombination, resulting in extended gain
lifetimes. By integrating the QSs with a photonic crystal (PhC) cavity with variable array
periods, we achieved tunable lasing with a near-record low lasing threshold. Our results
closely agree with theoretical model predictions, marking a significant advancement in the
development of colloidal nanocrystal lasers
Centralized and Distributed Algorithms in Wireless Sensor Networks
No full textWireless Sensor Networks (WSNs) play a significant role in various fields due to their ability
to collect data from remote or inaccessible locations without the need for wired connections.
There has been extensive research on designing topology and protocols for WSNs according
to numerous criteria such as energy consumption, cost-efficiency, availability, among others.
In this dissertation, we study some of the most fundamental problems of WNSs under both
centralized and distributed settings.
The first half of this dissertation, containing two parts, focuses on the algorithmic problems
of Directional Wireless Sensor Networks (DWSNs) under the centralized setting.
• In the first part, we study the Antenna Orientation (AO) problem and the Antenna
Orientation and Power Assignment (AOPA) problem of DWSNs where sensors are
equipped with k, (3 ≤ k ≤ 4) directional antennas with beam-width θ, (0 ≤ θ ≤ 2π).
For the AO problem, we design approximation algorithms for each case of k which
outperform the algorithms in previous works (Andersen and Ras, 2016; Tran and Huynh,
2018, 2020). Based on the algorithms for the AO problem, we propose approximation
algorithms for the AOPA problem which also surpass the algorithms in (Tran and
Huynh, 2018, 2020).
• The second part investigates the Directional Target Coverage and Connectivity (DCTC)
problem in DWSNs, which seeks a deployment of directional sensors with a minimum
number of nodes, such that all the nodes cover a set P of targets in the 2D plane and
form a connected communication network. Each sensor is equipped with a directional
sensing unit having beam-width θs, 0 ≤ θs < π and a directional communication antenna
with beam-width θc, θc ≥ π
2 . Each sensing unit and communication antenna has a
sensing range rs and communication range rc, respectively. The algorithm in this work
outputs a deployment of at most 3.5 times the optimal number of nodes, which improves
upon the trivial bound of 2π/π
2 = 4 when switching from omni-directional to directional
sensors.
The second half containing two parts, focuses on the design and analysis of algorithms in
WSNs under the distributed setting.
• The first part discusses the Minimum Dominating Set (DS) and Minimum Connected
Dominating Set (CDS) Construction of WSNs under the sleeping model (Chatterjee
et al., 2020). Unlike traditional WSNs where nodes need to be awake all the time, in
the sleeping model, nodes can sleep to save energy. The objective of algorithms for
the sleeping model is to minimize the time that nodes are awake. Our work proposes
awake-efficient algorithms to construct DS and CDS of small sizes.
• In the second part, we study the labeling schemes used used in distributed data
communication protocols in WSNs. Our work proposes an improvement from O(log ∆)
to O( log n) on the labeling size for the Convergecast protocol. Our results are the first
for labeling schemes for Broadcast and Data Aggregation in geometric graphs, and our
algorithms also achieve nearly optimal solutions. Finally, we extend our results to the
SINR interference model, which is the first results for labeling schemes in this model
3D Printing of Piezoresistive Flex Sensors for Soft Robotic Grippers and Strain Sensing
No full textFlexible piezoresistive sensors also referred to as strain gauges or flex sensors are extensively used
in soft robotics to detect strain due to their inherent flexibility. They change their resistance in
response to strain. This change in resistance can be used to detect bending in soft robots. 3D
printing based on filament extrusion (or Fused Filament Fabrication/ Fused Deposition Modeling)
of conductive polymers proffers the ability to fabricate inexpensive flex sensors quickly. There
seems to be a dearth of literature that characterizes FFF printed flex sensors by cyclic bending.
The effect of fabrication parameters, substrate materials, and sensing element geometry on the
cyclic bending response of the sensor needs to be assessed. In this thesis, a voltage divider circuit
is used for data acquisition and characterization experiments are conducted to determine
parameters for fabricating flex sensors with high repeatability (with a coefficient of variation less
than 2% which is lower than any other literature reviewed), good sensitivity (with a gauge factor
of over 6.5, more than three times higher than conventional metal strain gauges), and longevity
greater than a hundred thousand cycles. The longevity of the flex sensor is experimentally
demonstrated. The flex sensors manufactured are compared with additively manufactured strain
sensors in literature. To demonstrate their applications the flex sensors are incorporated in a soft
silicone robotic gripper and TPU finger in which they successfully detect the bending angle. The
temperature sensitivity of the sensor is also demonstrated. These sensors may have a potential for
sensing in several other soft structures that need flexibility as a prime factor