Engineering Commons^™

Online Hyperparameter Tuning For Llm Optimization, Ethan Lin, Nathan Yu, Jeromy Chang

Computer Science and Engineering Senior Theses

Large Language Models (LLMs) are becoming increasingly popular in modern society. However, despite their popularity, the deployment of LLMs in real-world scenarios is extremely challenging due to substantial computational costs and memory constraints. Edge devices, like smartphones and IoT devices, lack resources needed to run these models locally, instead offloading computations for cloud computing. Cloud computing requires users to send their data over the internet leading to numerous privacy and security concerns. In some domains, such as health and finances, sending such sensitive information is not an option. Existing solutions to compress or increase inference speed include Small Language Models …

The Dpe Shortage And How To Fix It, Paul Dye

Honors Theses

The following content delves into the current issues surrounding the FAA’s designated pilot examiner (DPE) system, which, while functional, is increasingly strained by growing demand and outdated policies. Through firsthand experience, survey data, and analysis of current practices, the paper emphasizes some of the most pressing problems, including scheduling inefficiencies, inconsistent examiner expectations, manual eligibility checks despite the widespread use of digital logbooks, and a lack of meaningful oversight or feedback mechanisms for both DPEs and CFIs.

A major focus is the outdated scheduling process, which remains largely decentralized and prone to double-booking, cancellations, and long wait times, especially for …

Intelligent Intrusion Detection In Clustered Wireless Sensor Networks: A Dynamic Clustering And Machine Learning-Based Approach, Abdullah R. Abdulwahhab, Mohd Fadzli Mohd Salleh, Muhammad Firdaus Akb, Mohammed Najm Abdullah

Iraqi Journal for Computer Science and Mathematics

Traditional Intrusion Detection Systems (IDS) designed for more conventional network infrastructures are often ill-equipped to handle the unique challenges WSNs pose, leading to significant gaps in security and resilience. This paper introduces an Intelligent Intrusion Detection System (IIDS) explicitly tailored for clustered WSNs to address these critical challenges. The proposed IIDS integrates dynamic clustering with advanced machine learning algorithms to create a robust and adaptive security solution capable of real-time threat detection and mitigation. The dynamic clustering mechanism is designed to continuously monitor and respond to changes in sensor node network topology and energy levels, ensuring that energy consumption is …

Non-Overlapping Patch-Based Pre-Trained Cnn For Breast Cancer Classification, Lamyaa Sabeeh Ashour, Ahmed Abed Mohammed, Mustafa M. Abd Zaid, Putra Sumari, Ahmed Kateb Jumaah Al-Nussairi, Sura Abdulateef Al-Shammari, Sarah Thabit Abdulmunem

Iraqi Journal for Computer Science and Mathematics

Breast cancer (BC) significantly impacts women's mortality rates and requires early detection to improve survival chances and enable appropriate treatment. Thus, a computer-aided system with high performance can speed up this process. A convolutional neural network (CNN) is considered sensitive to insufficient, noisy data. It cannot achieve high performance, however, restricted access to high-quality medical data, stemming from stringent confidentiality and privacy issues, is a considerable obstacle to the successful training of deep learning models. The current study aims to develop a remarkable, influential model for BC classification whilst considering modern pre-trained models ResNet50, AlexNet, InceptionV3 and VGG16 for extracting …

Retracted: Navigating The Complexities And Artificial Intelligence Of Internet Of Things Security Claims, Tamara Saad Mohamed, Saad Mohammed Khalifah

Iraqi Journal for Computer Science and Mathematics

The term ``Internet of Things'' (IoT) describes a system that allows everyday objects to communicate with one another and with controlled systems, servers, and other linked devices through a variety of connectivity constructions by means of embedded software, sensor technology, electronics, and connections. Data from the Internet of Things (IoT) will be sent to the servers over the internet from a variety of sensors, nods, and collectors. Governments, medical facilities, consumers, and corporations all make use of IoT devices. approximately, More than 65 billion Internet of Things devices are expected to be in operation by 2024. The proliferation of IoT …

Topological Indices For The Resize Graph Of (G2(3)), Manar Musab Ftekhan, Ali Abd Aubad

Iraqi Journal for Computer Science and Mathematics

Indexes of topological play a crucial role in mathematical chemistry and network theory, providing valuable insights into the structural properties of graphs. In this study, we investigate the Resize graph of G₂(3), a significant algebraic structure arising from the exceptional Lie group (G₂) over the finite field F₃. We compute several well-known topological indices, including the Zagreb indices, Wiener index, and Randić index, to analyze the graph's connectivity and complexity. Our results reveal intricate relationships between the algebraic structure of G₂(3) and its graphical properties, offering a deeper understanding of its combinatorial …

Machine Learning Algorithms To Detect Cyber-Attack In The Internet Of Things Platform, Maad Kamal Al-Anni, Rafah M. Almuttairi, Ammar A. Al-Hamadani, Khamis A. Zidan, Husam Ibrahiem Husain Alsaadi, Ghaidaa A. Al-Sultany

Iraqi Journal for Computer Science and Mathematics

In response to escalating cyber threats in Internet of Things (IoT) networks, this research investigates several traffic classification techniques by using benchmark datasets, such as CICIoT2023. This dataset was exclusively used to test the effectiveness of the proposed system, which aims to enhance the data processing for machine learning algorithms amid increasingly complex cyber threats. To prepare the naturally unstructured and raw dataset for analysis, Linear Discriminant Analysis (LDA) is applied for dimensionality reduction. The processed data and attributes are then fed into the proposed Fuzzy-Integrated Relevance Vector Machine Classifier (FIRVM), which is implemented in Python 3.10⁺ using the …

Multi Trajectory Guided Model Inversion Attacks, Indrajit Nandi

Master’s Dissertations

Recent advancements in deep learning have brought significant concerns regarding the privacy of training data due to overfitting and memorization, as well as a lack of defense mechanisms. In model inversion attacks, where Attackers aim to reconstruct original private training samples (i.e images) just by using the DNN model’s last layer output. Traditional black-box MI attacks face three key challenges: (1) Non-convex optimization landscapes often trap reconstructions in poor local minima, degrading output quality. (2) Black-box scenarios require excessive queries to approximate gradients, raising detection risks. (3) Unconstrained pixel-space optimization generates unrealistic artifacts since the inverse mapping lacks natural image …

Highly Adaptable Dendrimer Gel Nanoparticles With Dual Targeting Of Upar And Ribonucleotide Reductase R2 For Better Retention And Improved Therapeutic Outcomes In Triple-Negative Breast Cancer, Hsin Yin Chuang, Da Huang, Lin Qi, Vidit Singh, Anna Chernatynskaya, Yue-Wern Huang, Hu Yang

Biological Sciences Faculty Research & Creative Works

Triple-negative breast cancer (TNBC) accounts for approximately 15% of breast cancers and lacks estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2), rendering it unresponsive to hormonal or anti-HER2 therapies. Due to its poor prognosis and limited treatment options, there is an urgent need for targeted therapies. In this study, we developed highly adaptable polyamidoamine (PAMAM) dendrimer-based gel nanoparticles with dual-targeting capabilities against urokinase-type plasminogen activator receptor (uPAR) and ribonucleotide reductase R2 (R2). These nanoparticles were designed to target both TNBC cells and cancer-associated stromal cells by leveraging uPA-uPAR interactions and delivering the antisense oligonucleotide …

Measurement Of Moisture Levels In Oils And Lubricants Using A Novel Moisture Sensor, Aaron Swartz, Ronak Ali

Electrical and Computer Engineering Graduate Research

It is very challenging to measure moisture levels in oils. There is not a good method to measure it. Using the novel moisture sensor invented by Dr. Zhi David Chen, we tried various methods to measure the moisture levels in oils. The initial trials are to immerse the sensor chip into the oils to see any sensor reading changes with the change of moisture levels in oils. It was observed that the sensor reading was unstable even after immersion into the oil for two weeks. The idea for immersion of the sensor chip into the oils failed due to the …

Western Kentucky University Stormwater Utility Survey 2025, Warren Campbell

SEAS Faculty Publications

This survey provides data on 2147 U.S. and 82 Canadian stormwater utilities. The data is intended to provide useful data to communities that wish to enact stormwater utilities. Different fee systems are examined with the goal of avoiding legal exposure. Recently, the major source of pre-disaster mitigation funds was eliminated. States should consider state-wide stormwater utilities to make up for this loss.

Pathways To Efficient And Equitable Solutions For Large-Scale Routing Problems, Abhay Sobhanan

USF Tampa Graduate Theses and Dissertations

This dissertation addresses large-scale optimization problems in transportation emerging from hierarchical decision-making, equitable workload allocation, and innovative routing logistics. It presents three sets of contributions, each detailed in a separate chapter, and offers computational tools and insights to advance both the theory and practice of transportation systems.

The first work introduces a deep learning-enhanced genetic algorithm framework for solving the Hierarchical Vehicle Routing Problems (HVRPs). Traditional optimization approaches to such problems require extensive evaluation of multiple lower-level routing solutions and are computationally intensive. Our innovative method integrates a genetic algorithm with a pretrained graph neural network, which is trained on …

Monochromatic Empty Triangles In Two-Colored Point Sets, Adrish Paik

Master’s Dissertations

A key problem in combinatorial geometry is the identification of properties of a subset of a point set which has properties like convexity, monochromaticity, and emptiness. This thesis actual works on this sides of combinatorial geometry: efficiently finding empty monochromatic triangles in two-colored point sets on the plane. While earlier research focused on counting empty triangles in point set without any color, our work takes an optimal algorithm to explicitly detect them. By using geometric insights and visibility-based techniques. We then address a relaxed but equally compelling variant: monochromatic triangles containing at most one point of the opposite color. For …

Detection Of Fake News In Short Videos: A Multimodal Approach, Mona Kumari

Master’s Dissertations

The rise of generative models and affordable video editing tools has fueled the spread of fake and manipulated videos, undermining information reliabilityespecially on social media. Traditional detection methods, focused on single modalities like visual artifacts or text cues, often struggle with diverse, user-generated content. This dissertation presents a unified framework for fake video detection that integrates multimodal semantics, narrative structure, and propagation behavior. Visual, audio, text, and OCR features are extracted using pretrained models (CLIP, Wav2Vec2), and segment-level graphs are built to model narrative flow using Graph Attention Networks (GATv2Conv). User engagement dynamics are modeled via a bidirectional LSTM. A …

Santa Clara Radio Astronomy Project (Scrap) Iv - Data Processing, Noah Abe, Wesley Durbano

Computer Science and Engineering Senior Theses

The Santa Clara Radio Astronomy Project (SCRAP) IV: Data Processing is the fourth year of a multi-year project that aims to provide SCU with a fully functioning radio telescope in a costefficient manner. The project aims to provide this radio telescope as a hands-on tool that can be used not just for scientific discovery, but also for SCU academic use in a wide variety of different subjects, such as electrical engineering and physics. This year, our SCRAP team aimed to bolster the long-term operation of the telescope, which is a key feature it must have for it to be considered …

Resilience Oriented Dynamic Bayesian Network For Time Dependent Power Distribution System During Hurricanes, Kehkashan Fatima

Thesis/ Dissertation Defenses

In this dissertation, the overhead line failure due to extreme winds was investigated. The objective of the overhead line failure analysis was to evaluate the failure probability of line due to hurricane. The standard IEEE 33 bus test system was utilized to analyze the impact of hurricane wind speed intensity at distinct time and location on the grid. The test system was divided into four regions based on the number of buses and hurricane category (according to Saffire Simpson Hurricane Wind Scale). Regions 1, 2, and 3 cover 8 buses while region 4 covers 9 buses. Regions 1, 2, 3, …

Xmr: Extensible Model Representation, Jason Cisneros, Lucas Van Der Heijden

Computer Science and Engineering Senior Theses

In safety critical industries, such as aerospace, verification and validation are crucial steps in system development. Validation begins with decomposing a system into its requirements. These requirements are used to create the high level model of the system. The low level system that is created from this must then be verified. Manually ensuring that the system being verified exactly matches the validated design is tedious, time consuming, and error prone. To solve this, we have created an open source tool that can automatically generate the code from UML models. This enables going from the high level model to the low …

Deepfake Detection, Timothy Tong, Abem Lucas

Computer Science and Engineering Senior Theses

From elections, to wars, to people’s personal lives, deepfakes have become ever more present, blurring the distinction between reality and fiction. Many deep learning-based deepfake detection methods lack generalizability. They often overfit to deepfake generation techniques in their training distribution, and fail to detect deepfake generation techniques outside of their training distribution. One approach to solve the generalization problem has been to use Vision Language Models, such as the pretrained CLIP model, to extract features that generalize across different deepfake generation techniques, including diffusion and GAN images. However, most CLIP approaches only consider image level features for detection, such as …

Exploring Novel Methods For Real-Time Multi-Camera People Tracking In Machine Learning, Johan Kou, Connor Vallero

Computer Science and Engineering Senior Theses

Multi-camera people tracking is the process of tracking persons and their paths, continuously across different camera fields of view. It can help track suspects across large areas, and assist individuals in emergency situations. The current state-of-the-art (SOTA) method involves using geometric-consistent constraints, information on the appearance of subjects, and pose estimation for dealing with occlusion issues. This current SOTA works well, but is still lacking in its ability to handle occlusion and perform well in real-time applications on a network. With occlusion, the IDs assigned to persons can be accidentally swapped in high density areas, or places where there are …

Vintage Game Emulator, Andrew Katchour, Ashwin Raman

Computer Science and Engineering Senior Theses

This project explores the integration of artificial intelligence into retro-style arcade titles to enhance the gameplay, while still preserving the nostalgic aesthetic and overall mechanics of the titles. By using Pygame to explore clones of the existing video game titles, we modified the existing games to develop AI into the games using various AI algorithms. The system introduces adaptive enemy behavior that responds to the players actions and also introduces an additional player to some games that will play alongside the player as well. This helps to create a more engaging and unpredictable gaming experience. The AI logic is implemented …

An Ai-Driven Microfinance Platform To Enhance The Growth Of Small Businesses, Abdullah Naveed, Yunzhou Wang

Computer Science and Engineering Senior Theses

Microfinance provides financial services to businesses which do not have access to regular banking institutions. There have been many digital applications developed to support microfinance schemes globally. However, the applications currently on the market do not yet fully leverage technology features, which are already being widely used in traditional banking. In conjunction with the Miller Center of Social Entrepreneurship at Santa Clara University, we have developed an application utilizing the latest technologies to help small businesses and lenders in all stages of a microfinance project.

Variable Natural Language Translator For Video Games, Gabriel Bernini, Cole Propach

Computer Science and Engineering Senior Theses

In the field of video games, developers work to immerse players in a variety of ways. In one genre of video games, 2D exploration, or Metroidvania as its colloquially known, developers try to immerse players by letting them explore large worlds with the hope to overcome isolation. Often these worlds are sparse or dead worlds where the player is alone, but by exploring more of the world familiarity with the space is achieved and this isolation can be overcome. We hope to innovate in this genre by utilizing a new way to simulate and overcome this isolation, through language learning. …

Compliant Mechanisms For Scalable Systems And Precision Deployment, Jared Rodney Hunter

Theses and Dissertations

Compliant mechanisms offer advantages in precision applications where size, scaling, and environmental effects render traditional mechanisms inadequate or likely to fail. However, designing compliant mechanisms and integrating them into larger, more complex systems remains a challenging problem. This thesis presents work that addresses this issue with analytical models and experiments on flat springs, scalable compliant mechanisms, and origami arrays. Storing and releasing energy in tiny, flat compliant mechanisms opens the door for the design of innovative actuators and products in space-constrained situations. Scaling compliant mechanisms allows novel systems to be made that can operate across scales. By designing for scalability …

Towards A Robust Hydrologic Data Assimilation System For Hurricane-Induced River Flow Forecasting, Peyman Abbaszadeh, Fatemah Gholizadeh, Keyhan Gavahi, Hamid Moradkhani

Civil and Environmental Engineering Faculty Publications and Presentations

The Hybrid Ensemble and Variational Data Assimilation framework for Environmental Systems (HEAVEN) is a method developed to enhance hydrologic model predictions while accounting for different sources of uncertainties involved in various layers of model simulations. While the effectiveness of this data assimilation in forecasting streamflow has been proven in previous studies, its potential to improve flood forecasting during extreme events remains unexplored. This study aims to demonstrate this potential by employing HEAVEN to assimilate streamflow data from United States Geological Survey (USGS) stations into a conceptual hydrologic model to enhance its capability to forecast hurricane-induced floods across multiple locations within …

Human Gut Commensal Alistipes Timonensis Modulates The Host Lipidome And Delivers Anti-Inflammatory Outer Membrane Vesicles To Suppress Colitis In An Il10-Deficient Mouse Model, Ethan Older, Mary K. Mitchell, Andrew Campbell, Xiaoying Lian, Michael Madden, Yuzhen Wang, Lauren E. Van De Wal, Thelma Zaw, Brandon N. Vanderveen, Rodney Tatum, E. Angela Murphy, Yan-Hua Chen, Daping Fang, Melissa Ellermann, Jie Li

Faculty Publications

Correlative studies have linked human gut microbes to specific health conditions. Alistipes is one such microbial genus negatively linked to inflammatory bowel disease (IBD). However, the protective role of Alistipes in IBD is understudied, and the underlying molecular mechanisms remain unknown. In this study, colonization of Il10-deficient mice with Alistipes timonensis DSM 27924 delays colitis development. Colonization does not significantly alter the gut microbiome composition, but instead shifts the host plasma lipidome, increasing phosphatidic acids while decreasing triglycerides. Outer membrane vesicles (OMVs) derived from Alistipes are detected in the plasma of colonized mice, carrying potentially immunomodulatory metabolites into the …

Modeling And Simulation Of Tensegrity Structures With Continuous Cables, Austin R. Brown

Theses and Dissertations

This thesis presents a novel approach for modeling and simulating tensegrity structures with continuous cables. A Python-based simulation framework package was developed to address limitations in existing tensegrity modeling software, particularly the inability to model cables passing through multiple nodes. The framework utilizes virtual work methods to solve for equilibrium configurations of 2D, 2.5D and 3D tensegrity systems. Key features include: • A YAML-based input system for easy definition of tensegrity topologies. • Algorithms for handling continuous cables spanning multiple nodes. • Optimization techniques to find stable equilibrium states. • Capability to simulate 2.5D structures wrapped around cylindrical surfaces. • …

Design And Development Of Hip Implants For Longevity Through Integrating Advanced Groove Structures And Additive Manufacturing, Asarudheen Abdudeen

Thesis/ Dissertation Defenses

This doctoral research concerns the design and optimization of hip implants (HIs) to enhance performance, durability, and improve patient outcomes by addressing the key issues of wear, deformation, and stress distribution. Innovative surface groove designs have been introduced to both a solid and a hollow femoral head with the aim of reducing friction and wear. The addition of grooves, with both a hemispherical and rectangular cross-section, onto the femoral head reduced friction, as debris produced by the inner liner was trapped within the grooves, significantly reducing adhesive wear. A numerical simulation study compared the effects of surface modifications — both …

Reclaiming Data To Transform Anti-Trafficking Systems, Laura Hackney

SMU Human Trafficking Data Conference

AnnieCannons believes that the people closest to the problem are closest to the solution. Yet in the anti-trafficking movement, data about survivors is often collected without survivors. Co-founder and CEO Laura Hackney will discuss how reclaiming data can transform systems of care accountability, and prevention. She will share lessons learned from the development of ResourceFull, a digital platform led and co- designed with trafficking survivors to safely connect individuals to trustworthy services and track outcomes across fragmented care systems. There are ethical, technical, and human-centered challenges of collecting and of using data in a space where safety and trust are …

Evaluating The Physiological Effects Of A Spacesuit While Performing Cpr, Marielle Lenehan

Beyond: Undergraduate Research Journal

Limited research exists on the challenges of performing cardiopulmonary resuscitation (CPR) while wearing a spacesuit/pressure suit; but, during a space mission or following spacecraft egress, astronauts may need to perform Basic Life Support (BLS) on a crewmember. BLS consists of CPR and External Chest Compressions (ECC). In an emergency, astronauts may not have time to remove their intravehicular (IVA) spacesuits/pressure suits or wait for their heart rates to return to resting conditions and, therefore, would have to perform CPR while physically exerted. This study sought to evaluate the physiological exertion on rescuer performance while performing ECCs over 6 minutes while …

Llm Music Creation And Recommendation Applications, Curtis Robinson, Allan Yang, Thomas Liu

Electrical and Computer Engineering Senior Theses

This thesis explores the application of large language models (LLMs) in music analysis, creation, and interaction, focusing on their potential to reshape traditional workflows in the music domain. The study begins by contextualizing the role of artificial intelligence in music technology, particularly emphasizing the emergence of LLMs like GPT and their unique capabilities in multimodal and musical contexts. A comprehensive survey of current research and toolsets highlights both creative and analytical implementations, ranging from text-based music generation to music information retrieval. The core contribution is an experimental framework that integrates LLMs with music processing libraries, enabling novel interactions such as …