Computer Engineering Commons^™

Efficient Performance Recovery Of Pruned Llms For End Devices, Paul Vardhan Bethapudi

Master’s Dissertations

Large Language Models have shown strong performance across reasoning, language understanding, and generation tasks, but their computational and memory requirements make deployment on low resource devices difficult. This dissertation studies efficient performance recovery of pruned LLMs, focusing on whether a pruned model can regain useful task performance through parameter-efficient fine-tuning while preserving the benefits of compression. The work uses Llama-3.2-3B-Instruct as the dense reference model and applies activationaware Wanda pruning at multiple sparsity levels(20%, 30% and 50%). The pruned models are evaluated on language modeling and reasoning tasks using WikiText-2 perplexity and GSM8K (Grade School Math 8000 problems) accuracy. Adaptation …

Exploring Gene Regulatory Neural Network Biocomputing Of Bacteria, Adrian Merle Ratwatte

Dissertations and Doctoral Documents, University of Nebraska-Lincoln, 2023–

Artificial Intelligence (AI) has evolved from brain-inspired algorithms into a discipline that increasingly integrates with biological systems. While silicon-based platforms have advanced machine learning, they remain limited in energy efficiency and operation in environments beyond silicon. This motivates biological computing as an alternative, enabling efficient, resource-aware, and reconfigurable computing within living systems. This dissertation addresses these limitations by introducing a bacterial computing framework that models Gene Regulatory Networks (GRNs) as Gene Regulatory Neural Networks (GRNNs). The GRNN mirrors the structure and function of Artificial Neural Networks (ANNs) through gene-gene interactions across trans-omic layers, enabling natural, self-regulating information processing within living …

Operational Feasibility Of Reinforcement Learning For Vehicle Routing Under Heterogeneous Fleet Capacity Constraints, Freddy Giovanny Aviles Moreno

Electronic Theses and Dissertations

Reinforcement learning methods have demonstrated strong performance on vehicle routing benchmarks, yet their behavior under severe capacity constraints remains unexplored. This dissertation investigates whether PPO-based neural routing policies maintain operational viability when vehicle capacity is severely constrained, as occurs in resource-limited rural logistics settings.

Through controlled experiments on synthetic instances and validation on real-world rural healthcare networks in Florida, this research reveals a critical capacity threshold effect. Moderate capacity reductions from 40 to 20 produce negligible performance loss (4.1%), while severe reductions to capacity 10 trigger catastrophic failure with 243% degradation, manifested through degenerate single-customer routing patterns. Convergence analysis identifies …

En-Feat: An Effective Feature Selection Method Using Ensemble Approach, Sasank Nath, Dhruba Kumar Bhattacharyya

Mansoura Engineering Journal

Feature selection is a crucial step in machine learning and data preprocessing, significantly influencing model performance and interpretability. This paper presents a comprehensive study and contributions in the domain of feature selection by integrating traditional learning techniques with ensemble-based, proposing an effective approach. We propose a Mutual Information-based feature aggregation approach applied to union sets of features, aiming to derive an optimal subset of features that maximizes accuracy. Then, we employ an ensemble method that utilizes forward selection over union sets to identify the optimal feature subsets through sequential feature selection. Our ensemble-based feature selection method called En-feat, is evaluated …

Tinyvgg-Based Real-Time Degradation Classification For Adverse Driving Scenes Using A Newly Collected Iraqi Driving Dataset, Yousif N. Abbas, Matheel E. Abdulmunim, Nada H. Ali, Ismail A. Mageed

Journal of Soft Computing and Computer Applications

Environmental conditions such as low-light at night, fog scattering, glare artifacts, rain streaks, and rain smear distortions are significant issues of camera-based perception in Autonomous Vehicles (AVs). These degradations alter the statistics of the scene, mask structure, introduce non-uniform noise, and adversely affect downstream vision processes, including detection and tracking. To overcome this shortcoming, this paper presents a lightweight TinyVGG-based degradation classification system that runs in real time. The network extracts discriminative spatial features with hierarchical convolutional encoding and projects them to a lower-dimensional semantic representation with fully connected layers and a multi-class predictor based on SoftMax. In addition, a …

Improving Approach Of Evolutionary Strategies For Clustering Technique Enhancement, Duaa Mahde Saleh, Hasanen S. Abdullah, Ahmad Zamsuri

Journal of Soft Computing and Computer Applications

The existence of the information has been the essential aspect of the whole society. Information is concentrated in all forms to be effectively utilized. Clustering — an unsupervised learning technique. It is based on data similarity that gives rise to issues in collection, challenges and instability in data structure. It proposes an advanced evolutionary method by combining two approaches. Firstly, it adopts the evolutionary approach and integrates the advantages between two methods to design one. Among them are Differential Evolution (DE) and Genetic Algorithm (GA), Evolutionary Strategy (ES) and Genetic Programming (GP), and Evolutionary Programming (EP) and Particle Swarm Optimization …

A Comprehensive Review Of 1d Deep Learning Approaches In Facial Analysis: Face Recognition, Landmark Detection, And Mesh Modeling, Duaa J. Al Hammami, Rehab F. Hassan

Journal of Soft Computing and Computer Applications

Facial Analysis has progressed rapidly with deep learning and its 2D image-based models, especially Convolutional Neural Networks (CNNs), which have been the most popular methods. In recent years, 1D deep learning models have gained traction in the search for efficient solutions for face recognition, facial landmark detection, and 3D face mesh modeling. 1D models encode the facial structure as sequences, curves, or temporal signals, resulting in high computational efficiency, a small memory footprint, and good interpretability, making them well-suited for real-time and edge devices. This review is a step-by-step, organized exploration of 1D deep learning analysis of the face, its …

Comparative Study On Throughput Optimization In Nfv: Traditional Dissemination Techniques Vs. Swarm Intelligence Approaches, Sanaa Salih Alwan, Asia Ali Salman, Wulfrano Arturo Luna Ramírez

Journal of Soft Computing and Computer Applications

Network Functions Virtualization (NFV) modernizes networks by replacing hardware with software, creating a more flexible network architecture and offering flexibility in dynamic network environments. This foundational technology is essential for creating the networks of the future, including the Internet of Things (IoT) and cellular services. NFV does provide flexibility, but it struggles to maintain system throughput during high traffic loads while achieving high resource utilization efficiency and dynamic packet routing. The problem lies in the fact that traditional request distribution mechanisms, such as flooding and gossip, fail to operate efficiently in complex network topologies (scale-free networks), leading to: (a) random …

A Comprehensive Review Of The A* Algorithm: Evolution, Applications, And Future Trends In Path Planning, Saleel H. Abood, Hussein M. H. Al-Khafaji, Mohanned M. H. Al-Khafaji

Journal of Soft Computing and Computer Applications

Despite being a fundamental problem to autonomous robotics and intelligent navigation systems, path planning is still a challenge. The A^* algorithm is often used among search-based techniques for optimal search performance, as it's a tradeoff of computation. The above techniques have been developed for various applications as many versions of A^* Dynamic A^* (D^*), D^* Lite, Hybrid A^*, and Anytime A^* are suggested to deal with dynamic environments, real-time constraints, and kinematic restrictions. This paper comprehensively and structurally reviews the A^* algorithm and its major extensions, encompassing historical development, methodological …

Skin Lesion Classification Using Cnn Model And Augmented Dataset, Mohammed Nawzad Mohammed-Ramzi, Aso M. Aladdin

Journal of Soft Computing and Computer Applications

Skin cancer is a deadly disease. Skin lesion classification is a critical challenge due to its prevalent and deadly nature. Skin lesions are difficult for dermatologists to detect using eye examination, which is time-consuming and variable. A deep learning model of skin lesions classification has been proposed using a Convolutional Neural Network (CNN) trained on the HAM10000 dataset of 10,015 dermatoscopies. To improve resilience and address the dataset's extreme class imbalance, data augmentation techniques such as geometric transformations, brightness/contrast adjustments, blurring, noise addition, histogram equalization, color space alterations, and elastic deformations are used. With a carefully balanced 10% test set, …

Development And Standardization Of Teds Actuator Templates Under Ieee 1451 Framework, Jim Kang

Theses and Dissertations

The IEEE 1451 standard is a family of standards that defines a framework for smart transducers, including both sensors and actuators, to support consistent, interoperable, and cost-effective integration across diverse applications. However, the current IEEE 1451.4 standard templates only define the method for encoding Transducer Electronic Data Sheet (TEDS) information for a broad range of sensor types and applications; they do not address actuator TEDS. Specific types of sensors and actuators are being developed to assess underground environmental conditions in cold regions with widespread permafrost. Evaluating subsurface conditions before construction can help prevent high construction expenses for structures built on …

Robust Real-Time Uav Target Tracking With Onboard Vision-Based Yaw Control, Rylan Malarchick, Jose Castelblanco, Enrique Amaya, Carmen Dimario, Graysen Brinkman, Chirag Kumar, Kiwon Yoon, Sajid Berhane

Beyond: Undergraduate Research Journal

Autonomous tracking of agile unmanned aerial vehicles (UAVs) presents significant challenges for real-time perception and control systems. This work presents AIRHOUND (Autonomous Intelligent Rotorcraft for Hostile Object Unified Navigation and Detection), a UAV platform implementing vision-based yaw tracking through a modular ROS2 software architecture. The system employs YOLOv8 object detection optimized with NVIDIA TensorRT for embedded deployment on an NVIDIA Jetson Orin companion computer. Detected targets are processed through a geometric tracking module that converts pixel coordinates to angular yaw errors using pinhole camera intrinsics, with a proportional controller generating rate-limited yaw commands. These commands are streamed to a PX4 …

Bias Before Generation: Attention-Based Preemptive Fairness Signals In Large Language Models, Aniket Das

Master’s Dissertations

Warning: This paper includes examples of language that may be perceived as inappropriate or offensive. Large language models (LLMs) are known to propagate social biases embedded in their training corpora, producing outputs that disproportionately disadvantage individuals based on sensitive attributes such as gender, religion, race, sexual orientation and nationality. Existing mitigation strategies are either computationally prohibitive, require access to model parameters, or apply corrections only after biased content has already been generated. This work addresses a different question: can the model’s own internal attention dynamics, observed at inference time, serve as a reliable early-warning signal for bias, enabling intervention before …

Air-Writing Recognition, Gaurang Shukla

Master’s Dissertations

Air-writing is the act of tracing characters or words in free space with a fingertip, recorded by a camera, giving a touch-free input modality for smart displays, augmented and virtual reality, and assistive interfaces. It is difficult because the finger never lifts: connecting strokes join adjacent letters with no pen-up signal to mark boundaries, and the same word varies widely in scale, position, and slant across writers. The WiTA benchmark of Kim et al. provides a large, person-disjoint dataset and a baseline that treats each clip as RGB video, recognised by a spatio-temporal 3D residual network trained with a CTC …

Jalzap: An Agile Hybrid For Ai-Accelerated Software Development, Pouya Nouri

University Honors Theses

The rapid adoption of generative AI tools allows software teams to quickly code applications, but it comes at a cost: high scope volatility, technical debt, and unrealistic expectations, which break traditional Agile frameworks. This thesis introduces JALZAP, a lightweight, hybrid Agile framework designed to serve small, high-agility teams facing compressed timelines and unpredictable schedules. This framework was evaluated over 16 weeks through a Portland State University Capstone project working for a pre-seed startup sponsor, where a six-person team built Flowmind: an AI-powered iOS task management app meant to serve users with neurodevelopmental disorders like ADD/ADHD. JALZAP implements structural boundaries, including …

Semantic Directing: A Human-Ai Workflow For Interative Virtual Cinematography, Lejie Liu

Dartmouth College Master’s Theses

This thesis presents AI Director, a Unity-based prototype for semantic directing in virtual cinematography. The system enables creators to describe cinematic intent in natural language, receive editable shot-strategy plans generated by a large language model, and refine specific shots through text and visual references. By separating semantic planning from deterministic camera execution, the workflow keeps user authorship centered on shot intent while making the resulting camera plans inspectable, editable, and executable in real time. This work explores semantic directing as a human-AI workflow for iterative cinematic authoring in dynamic 3D scenes.

Universal Communications Module: Converting Ieee2030.5 Xml Files To Ansi/Cta-2045, Wyatt Bilodeaux

University Honors Theses

The goal of this thesis is to discuss the method and organization of the Universal Communication module capstone project. This project was part of the Portland State University Electrical engineering degree program. While this thesis is part of the University honors curriculum. This thesis goes into detail about team structure and how a 9 person engineering capstone project was managed and organized. It looks at how we broke down the project into sections to be able to properly distribute our resources. This was one of the earliest and most important challenges that this capstone team had to tackle. This was …

Interactive Social Game Design For Sustainability Learning, Kayla Mattson, Kayla Huffman, Megha Thottam, Shreeya Koritala

Computer Science and Engineering Senior Theses

Our Roots is a community-centered, web-based social game designed to improve sustainability education by turning everyday environmental actions into shared, visible progress. While many existing tools provide information about sustainability, they often focus on individual tracking and short-term motivation, which can lead to disengagement over time. This project addresses that gap by reframing sustainability as a social practice: users log real-world sustainable behaviors, earn points, and see those actions represented as growth in both personal and community digital gardens. By combining friendly competition, peer visibility, and collective outcomes, the platform encourages users to view sustainability not as an isolated task, …

A Novel Design-For-Test Flow Using Librelane, Mohamed E. Gaber

Theses and Dissertations

Design-for-testing is a crucial element of application-specific integrated circuit design, yet, in the nascent open-source electronic design automation scene, the solutions for it are sorely limited. Design-for-test-enabled chips allow defects to be caught early on in the manufacturing process, avoiding incurring huge costs if hardware with defective chips is shipped to equipment manufacturers or, worse, end-users. Yet, the current open-source solutions rely on a brute-force utility that does not scale and negatively impacts the design by holding design-for-testing features as co-equal with the design’s regular operation, and a nominally layout-aware solution that does not holistically integrate into a larger flow. …

Bouillean - Ai Recipe Model, George Mouratoff, Hanol Yang, Helen Wang, Sebastian Verbinski

Computer Science and Engineering Senior Theses

AI generated food recipes are the second most desired AI innovation, however relatively few commercial systems exist to fill this niche. This is because while the solution is highly desired, cooking shouldn’t be approached with the same method by most LLMs and standard AIs. The leading models struggle to solve the issue of generating a semantically clear recipe while maintaining memory of the recipe status between steps. This is before we can even attempt to solve the flavor problem.

Cooking represents a unique challenge for AI, since it requires minimal loss in memory alongside already high quality text generation making …

Ai In Precision Medicine: Redefining Blindness Treatment Via Retinal Stem Cell Differentiation, Calissa Leong, Tiffany Nguyen

Interdisciplinary Design Senior Theses

Age Macular Degeneration (AMD) is currently the leading cause of blindness, driven by the degradation of Retinal Pigment Epithelial (RPE) cells. Stem-cell based therapies, particularly those involving induced pluripotent stem cells (iPSCs), typically consists of differentiating stem cells into RPE cells in vitro and transplanting them into the subretinal space. Our goal is to improve the accuracy of differentiation into RPE cells to better promote retinal regeneration. Traditional methods to alleviate symptoms of AMD is through therapeutic medications to slow disease progression, though stem cell therapy has emerged as a promising alternative. However, despite its potential, stem cell therapy presents …

Fisheye Camera Object Detection For Traffic Surveillance, Hugo Yang, Robert Luo

Computer Science and Engineering Senior Theses

Effective traffic management is hampered by the physical limitations of traditional pinhole cameras, whose narrow fields of view create blind spots that require costly multi-camera infrastructure to overcome. Fisheye cameras offer a wide-angle alternative capable of single-camera intersection coverage, but the radial distortion inherent in their optics renders standard object detection algorithms unreliable, limiting their practical deployment in traffic surveillance.

This thesis investigates specialized object detection pipelines for fisheye traffic imagery by reproducing and building upon two top-ranked submissions from the 2025 AI City Challenge Track 4: the VNPT and SKKU-AutoLab pipelines. Both were successfully reconstructed outside their original containerized …

Ml Models For Child-Parent Speech Diarization, Kenneth Kang

Computer Science and Engineering Senior Theses

This senior design thesis presents a machine learning research pipeline for child-parent and child-directed speech diarization. Speech diarization is the task of determining who spoke and when in an audio recording. Although this problem is well studied in adult meeting, broadcast, and conversational speech, child-parent audio is a more uneven setting. Children may speak in short turns, change pitch quickly, vocalize without producing clear words, interrupt adults, overlap with adult speech, cry, laugh, or code switch. These behaviors are normal in real family interaction, but they can confuse diarization systems that were mostly designed and evaluated on adult speech.

The …

Trends In Non-Profit Cybersecurity: Analyzing Three Years Of Incident Data From The Npcir, Stanley Mierzwa, Joanna Paliszkiewicz, Edyta Skarzyńska

Center for Cybersecurity

This study analyzes cyberattack trends targeting non-profit organizations using longitudinal data collected over a three-year period within the Non-Profit Cybersecurity Incident Repository (NPCIR). Developed through a National Security Agency Center of Academic Excellence in Cyber Defense (NSA CAE-CD) designated center, the NPCIR applies an open-source intelligence (OSINT) methodology to systematically document cybersecurity incidents affecting the global non-profit sector. This study examines attack types, threat actor characteristics, sectoral distribution, and cybersecurity impacts using the Confidentiality–Integrity–Availability (CIA) triad framework. The results indicate that availability-related incidents, particularly ransomware and distributed denial-of-service (DDoS) attacks, constitute the most prevalent threats, while confidentiality breaches remain highly …

Implementation Of A Mixed Precision Iterative Solver On An Amd Apu, Saeyeon Kim, Henrik Evers

Computer Science and Engineering Senior Theses

Solving systems of linear equations is a core operation in high-performance computing and many scientific and engineering applications. One of the most widely used direct methods for solving such systems is LU factorization. Mixed-precision methods accelerate these solvers by using low-precision LU factorization with iterative refinement, improving performance and energy efficiency while preserving high numerical accuracy. AMD Accelerated Processing Units (APUs), which integrate CPU and NPU resources, provide an opportunity to explore this approach on emerging heterogeneous architectures. However, effective NPU utilization remains challenging due to workload partitioning, buffer management, data-movement overhead, and software tooling limitations.

This project implements a …

Broncosgo: Driving Student Event Participation Via Human-Ai Collaborative Event Curation, Katherine Carter, Ethan Sychangco

Computer Science and Engineering Senior Theses

College campuses rely heavily on email communication for event advertisement. However, due to the sheer number of emails any student may receive, the individual effectiveness of each advertisement decreases. A better alternative for event promotion was needed. How could we introduce students to events they would not usually find on their own, without continuing to fill up their email inbox? And how would information on all of those events be collected and maintained without requiring organizations to jump through extra hoops? Our project, BroncosGo, solves this issue by maintaining a database of campus events which is populated with information provided …

Club Finance Management Application, India Bell, Lydia Martin, Sola Kawahara, Zachariah Wang

Computer Science and Engineering Senior Theses

There are over 160 Registered Student Organizations (RSOs) at SCU as of 2023 [6]. There is currently no centralized system that RSO treasurers use to track and manage club finances. As a result, many organizations rely on fragmented methods such as spreadsheets, shared documents, email chains, and manual record keeping to manage budgets, reimbursements, and financial reports. These disconnected systems often lead to inconsistent organization, communication difficulties, and challenges during leadership transitions between club officers. This project proposes the development of a web-based Club Finance Management Application, Kaikei, designed specifically for Santa Clara University student organizations. The application centralizes financial …

Iajes Website Redesign, Justin Fan, Nicolas Gibson, Isabelle Hu, Anisha Malani

Computer Science and Engineering Senior Theses

The International Association of Jesuit Engineering and Science Schools (IAJES) connects more than sixty Jesuit institutions worldwide to promote collaboration and shared learning in STEM research. Santa Clara University is one of more than sixty member universities whose faculty and staff participate in conferences and task forces to address priority issues identified by the Society of Jesus. IAJES’ existing website, built on a static no-code platform, limited the organization’s ability to manage content, support member interaction, and scale with its growing global network. Users faced unclear navigation options, a basic design, and no effective tools for managing roles, resources, or …

Bilingual Math For Kindergartners (Biki), Anna Aldrin, Caroline Tapia, Lillian Le

Computer Science and Engineering Senior Theses

For many children from bilingual households, kindergarten represents a critical, learning benchmark where the introduction of fundamental math concepts coincides with their first significant exposure to the English language. Traditional educational methods often fail to support English language learners, leading to a disconnect between their understanding of math in the home and its application in the classroom. This linguistic barrier frequently results in reduced engagement, diminished confidence, and an unstable educational foundation that can discourage students from pursuing STEM careers later in life. While existing platforms like Khan Academy and Duolingo address either language acquisition or mathematical proficiency independently, they …

Bridgeide, Maya Srimal, Jeshwin Prince, Arnav Gupta, Will Zumbolo

Computer Science and Engineering Senior Theses

Anyone who has done a group project knows the pattern: the first meeting burns an hour on logistics—which document to use, who has access, why the formatting broke on one laptop—before real work begins. In computer science courses that gap is worse. Students must install the right compiler, runtime, and libraries on their own machines; a four-person team on four different laptops can lose the first week to setup alone. Instructors who want to review submissions often download every zip from the learning management system and run each one locally. Existing tools solve pieces of this puzzle—real-time document editing, browser-based …