Treasures @ UT Dallas
Welcome to Treasures @ UT Dallas Institutional Repository, established in 2010. Treasures is a resource for our community to showcase, organize, share, and preserve research and scholarship in an Open Access repository.
Communities in Treasures
Select a community to browse its collections.
Recent Submissions
Transfer Learning and Uncertainty Quantification in Natural Language Processing for Political Science and Cyber Security
(2023-08) Hu, Yibo; Khan, Latifur; Makris, Yiorgos; Ouyang, Jessica; Brandt, Patrick T.; Du, Xinya
Recent advancements in Natural Language Processing (NLP) driven by pretrained language
models have revolutionized various fields reliant on large-scale text-based research through
transfer learning. This dissertation presents efficient, reliable computational NLP applications to address real-world challenges, with a focus on political science, cyber security, and uncertainty quantification.
The dissertation begins with interdisciplinary research in political science, where advanced
NLP models are developed to track and analyze dynamics related to global political conflict.
The creation of ConfliBERT, the first domain-specific sociological language model, enables
improved performance on 18 downstream tasks, particularly in scenarios with limited data
availability. Moreover, by leveraging transfer learning and existing expert knowledge, specific tasks such as political event extraction and classification are further optimized. One
approach called Confli-T5 is a text generation model that augments labeled data by in-
corporating achievable templates derived from political science knowledge bases. Another
technique introduced is the Zero-Shot fine-grained relation classification model for PLOVER
ontology (ZSP), which eliminates the need for labeled data by relying solely on an annotation
codebook to classify intricate interactions between political actors. These strategies combine
the power of transfer learning with domain-specific expertise to reduce the dependence on
extensive labeled data, making them valuable tools in the field.
In the field of cyber security, text generation techniques are employed for cyber deception, generating multiple fake versions of critical documents to deter malicious intrusion.
A context-aware model called Fake Document Infilling (FDI) addresses the limitations of
existing approaches by considering contextual awareness. FDI produces highly believable
fake documents, protecting critical information and deceiving adversaries effectively.
Finally, uncertainty quantification techniques are explored to enhance the reliability of NLP
models in such interdisciplinary or cross-domain applications. A novel model, BERT-ENN,
employees evidential theory to quantify multidimensional uncertainty in the data and calibrate uncertainty estimation in text classifiers. This approach achieves state-of-the-art
out-of-distribution detection performance, thereby improving the reliability of NLP models.
Risk-based Motion Planning and Control for Robotic Systems
(2023-12) Safaoui, Sleiman; Summers, Tyler; Kang, Gu Eon; Spong, Mark W.; Koeln, Justin; Ruths, Justin; Vinod, Abraham P.
A robot autonomy stack usually consists of several modules that enable it to perceive the
environment and decide how to interact with it to achieve a desired task. At the heart of
this stack are the motion planning and control modules. The motion planning module is
generally responsible for decision making and generating a plan for the robot to follow, such
as determining how an autonomous car should drive around pedestrians and other vehicles.
The control module computes a finer sequence of control actions that can be issued to the
actuators to operate the robot.
One issue that plagues robot motion planning and control is the effect of uncertainty, of which
there are different types, on the system. This includes unknown and unmodeled disturbances
that affect the system such as noise, aerodynamics, or simplified dynamics models. However,
addressing these uncertainties is non-trivial and often requires a trade-off between accounting
for the uncertainty accurately and the tractability of solving the problems.
This dissertation develops risk-based solutions for a few robot motion planning and control
problems. The contributions of the dissertation are categorized into four main types.
The first part addresses control design with complex spatio-temporal requirements under uncertainty. An optimization-based control algorithm is designed to guarantee the completion
of the requirements when the robot dynamics are affected by process noise.
The second part addresses sampling-based motion planning under uncertainty. RRT*, a
famous motion planning algorithm in robotics, is considered and risk-aware variants of it are
developed to account for process and measurement noise affecting the robotic system.
The third part addresses a limitation of learning-based planning approaches with an application to multi-agent motion planning. A reinforcement learning (RL) framework is considered
for learning policies then an optimization-based module, called a safety filter, is proposed to
enforce collision avoidance as hard constraints, which learning algorithms cannot do. The
safety filter is designed to handle process, state, and measurement noise.
Finally, the fourth part addresses data-driven planning in dynamic and uncertain environments. This assumes that the robot has access to some future predictions of the obstacles in
the environment, such as where they may be in the next few seconds. A safety filter is then
developed using these sample predictions to plan a safe trajectory for the robot.
In several sections, uncertainties whose distribution is unknown, which is generally the
case, are considered and addressed using the concept of distributionally robust optimization (DRO) to develop solutions that guarantee safety or the successful completion of the
task despite the lack of knowledge of the underlying distribution.
Throughout, examples are provided to emphasize and clarify core concepts, and simulations and physical experiments are performed to demonstrate the efficacy of the developed
solutions.
Student Haiku Competition 2024
(Eugene McDermott Library, 2024) Aries De Joy Uy, Lester; Raghavaraiu, Nikhitha; Ramsey, Carson