Computer Engineering Commons^™

Deep Learning Techniques For Image Segmentation In Dermoscopic Skin Cancer Images, Norsang Lama

Doctoral Dissertations

"Melanoma is recognized as the most lethal type of skin cancer, responsible for a significant proportion of skin cancer-related deaths. However, early detection of melanoma is essential for successful treatment outcomes. Computer-aided skin cancer diagnosis tools can save lives by enabling earlier detection of skin cancer. Image segmentation is a crucial step in computer-aided diagnosis as it allows the detection of critical features or regions in an image. Thus, an accurate image segmentation method is necessary to create a more precise computer-aided diagnostic tool for skin cancer diagnosis. This dissertation includes investigating and developing deep learning techniques to improve image …

A Lightweight Machine-Learning Framework For Enhancing Security In Iot Blockchain Networks, Charles Connor Rawlins

Doctoral Dissertations

"Blockchain is one of the fastest technologies that rivals the Internet in terms of adoption speed. This security method is applicable to data-centric environments for validating data in the presence of faults. However, traditional blockchain implementation introduces bottlenecks with computationally intense security measures to prevent malicious spam and resolve conflicts. This dissertation explores a new direction for blockchain technology that allows limited nodes, like IoT devices, to make independent decisions with compressed knowledge of past blockchain history through the use of machine-learning for active decisions (or the first machine-intelligent blockchain protocol). Proposing to introduce machine-intelligence into the rapidly evolving paradigm …

Modeling And Control For Precision Robotic Machining, Patrick Bazzoli

Doctoral Dissertations

"Robots are used in a wide variety of manufacturing applications, but machining applications in which robots can excel are limited by their lower accuracy and stiffness relative to traditional CNC machines. This work is composed of two parts: one to evaluate a robot’s accuracy and one to compensate for the vibrations of the robot due to its lower stiffness.

In order to evaluate whether a robot has the necessary accuracy to perform a given machining task, Paper 1 discusses a novel Model Invalidation method. This methodology provides a statistical framework as well as a measurement strategy for determining if a …

Railroad Condition Monitoring Using Distributed Acoustic Sensing And Deep Learning Techniques, Md Arifur Rahman

College of Graduate Studies: Theses & Dissertations

Proper condition monitoring has been a major issue among railroad administrations since it might cause catastrophic dilemmas that lead to fatalities or damage to the infrastructure. Although various aspects of train safety have been conducted by scholars, in-motion monitoring detection of defect occurrence, cause, and severity is still a big concern. Hence extensive studies are still required to enhance the accuracy of inspection methods for railroad condition monitoring (CM). Distributed acoustic sensing (DAS) has been recognized as a promising method because of its sensing capabilities over long distances and for massive structures. As DAS produces large datasets, algorithms for precise …

Assessing Performance Optimization Strategies In Cloud-Native Environments Through Containerization And Orchestration Analysis, Daniel E. Ukene

College of Graduate Studies: Theses & Dissertations

This thesis comprises three distinct, yet interconnected studies addressing critical aspects of web infrastructure management. We begin by studying containerization via Docker and its impact on web server performance, focusing on Apache and Nginx hosted on virtualized environments. Through meticulous load testing and analysis, we provide insights into the comparative performance of these servers, adding users of this technology know which webservers to leverage when hosting their webservice along alongside the infrastructure to host it on. Next, we expand our focus to examine the performance of caching systems, namely Redis and Memcached, across traditional VMs and Docker containers. By comparing …

Neurosymbolic Value-Inspired Ai (Why, What, And How), Amit Sheth, Kaushik Roy

Publications

The rapid progression of Artificial Intelligence (AI) systems, facilitated by the advent of Large Language Models (LLMs), has resulted in their widespread application to provide human assistance across diverse industries. This trend has sparked significant discourse centered around the ever-increasing need for LLM-based AI systems to function among humans as part of human society, sharing human values, especially as these systems are deployed in high-stakes settings (e.g., healthcare, autonomous driving, etc.). Towards this end, neurosymbolic AI systems are attractive due to their potential to enable easy-tounderstand and interpretable interfaces for facilitating valuebased decision-making, by leveraging explicit representations of shared values. …

Expanding Australia's Defence Capabilities For Technological Asymmetric Advantage In Information, Cyber And Space In The Context Of Accelerating Regional Military Modernisation: A Systemic Design Approach, Pi-Shen Seet, Anton Klarin, Janice Jones, Michael N. Johnstone, Violetta Wilk, Stephanie Meek, Summer O'Brien

Research outputs 2022 to 2026

Introduction. The aim of the project was to conduct a systemic design study to evaluate Australia'sopportunities and barriers for achieving a technological advantage in light of regional military technological advancement. It focussed on the three domains of (1) cybersecurity technology, (2) information technology, and (3) space technology.

Research process. Employing a systemic design approach, the study first leveraged scientometric analysis, utilising informetric mapping software (VOSviewer) to evaluate emerging trends and their implications on defence capabilities. This approach facilitated a broader understanding of the interdisciplinary nature of defence technologies, identifying key areas for further exploration. The subsequent survey study, engaging 828 …

Multimodal Fusion For Audio-Image And Video Action Recognition, Muhammad B. Shaikh, Douglas Chai, Syed M. S. Islam, Naveed Akhtar

Research outputs 2022 to 2026

Multimodal Human Action Recognition (MHAR) is an important research topic in computer vision and event recognition fields. In this work, we address the problem of MHAR by developing a novel audio-image and video fusion-based deep learning framework that we call Multimodal Audio-Image and Video Action Recognizer (MAiVAR). We extract temporal information using image representations of audio signals and spatial information from video modality with the help of Convolutional Neutral Networks (CNN)-based feature extractors and fuse these features to recognize respective action classes. We apply a high-level weights assignment algorithm for improving audio-visual interaction and convergence. This proposed fusion-based framework utilizes …

Reinforcement Learning From Human Feedback For Ethically Robust Ai Decision-Making, Marco M. Plasencia

Honors Undergraduate Theses

The emergence of reinforcement learning from human feedback (RLHF) has made great strides toward giving AI decision-making the ability to learn from external human advice. In general, this machine learning technique is concerned with producing agents that learn to work toward optimizing and achieving some goal, advanced by interactions with the environment and feedback given in terms of a quantifiable reward. In the scope of this project, we seek to merge the intricate realms of AI robustness, ethical decision-making, and RLHF. With no way to truly quantify human values, human feedback is an essential bridge in the learning process, allowing …

Robotic Gas Source Localization And Distribution Mapping Via Deep Reinforcement Learning, Iliya Kulbaka

UNF Graduate Theses and Dissertations

This research aims to advance the fields of Gas Source Localization (GSL) and Gas Distribution Mapping (GDM) by developing deep reinforcement learning (DRL) methodologies suitable for complex, real-world environments. GSL and GDM are crucial for applications such as environmental monitoring, hazardous material detection, and search-and-rescue missions, where safe and efficient exploration is essential. Traditional methods often fall short in dynamic settings influenced by factors like wind and obstacles. To address these limitations, this study proposes novel neural network architectures and learning frameworks for adaptive exploration and mapping, integrating Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM) layers, and Deep Q-Networks …

A Memory Efficient Deep Recurrent Q-Learning Approach For Autonomous Wildfire Surveillance, Jeremy A. Cantor

UNF Graduate Theses and Dissertations

Previous literature demonstrates that autonomous UAVs (unmanned aerial vehicles) have the po- tential to be utilized for wildfire surveillance. This advanced technology empowers firefighters by providing them with critical information, thereby facilitating more informed decision-making processes. This thesis applies deep Q-learning techniques to the problem of control policy design under the objective that the UAVs collectively identify the maximum number of locations that are under fire, assuming the UAVs can share their observations. The prohibitively large state space underlying the control policy motivates a neural network approximation, but prior work used only convolutional layers to extract spatial fire information from …

Transformer-Enabled Deep Reinforcement Learning For Coverage Path Planning, Daniel B. Tiu

UNF Graduate Theses and Dissertations

Coverage path planning (CPP) is the problem of covering all points in an environment and is a well-researched topic in robotics due to its sheer practical relevance. This paper investigates such an offline CPP problem where the primary objective is to minimize the path length to achieve complete coverage. Furthermore, the literature suggests that taking turns leads to a higher energy use than going straight. To this end, we design a novel objective function that aims to minimize the number of turns as well. We have proposed a deep reinforcement learning (DRL)-based framework that uses a Transformer model. Unlike state-of-the-art …

Explainable Automated Inconsistency Detection In Biomedical And Health Literature, Prajwol Lamichhane

UNF Graduate Theses and Dissertations

Given the exponential growth of scientific information online, researchers often face the daunting task of detecting contradictory statements on crucial health topics. This work develops a comprehensive pipeline for automated contradiction detection that integrates an Information Retrieval (IR) system, machine learning classifiers, and Explainable AI (XAI). The Information Retrieval system is tailored for biomedical data and comprises a datastore, syntactic, and semantic components. Users can input queries, initiating a pipeline that identifies top documents through syntactic analysis and refines results via semantic examination for relevant research claims. Employing a diverse range of Large Language Models such as pre-trained Distil-BERT, BioBERT, …

Large Language Models For Multimodal User Interaction In A Virtual Environment, Ahmed A. Sayed

UNF Graduate Theses and Dissertations

Virtual Reality (VR) is increasingly popular, but many barriers exist for individuals with little experience in coding, 3D modeling, or creating their own virtual experiences. The current tools used for content creation are often viewed as complex or frustrating, and they exhibit a steep learning curve. This problem presents an opportunity to develop tools incorporating Natural User Interfaces that better support end users. One such possible tool to assist users is Large Language Models (LLM), which can, extract a user's intention through speech or text. We posit how using LLMs can better support novice and expert developers alike, and using …

Finops-Driven Cloud Optimization Models For Enterprise Applications, Manikantha Varaprasad Inakollu

Computer Science and Engineering Faculty Publications

Cloud computing has revolutionized enterprise IT infrastructure, yet escalating costs and resource inefficiencies threaten to undermine these benefits. This research examines FinOps-driven optimization models that enable organizations to balance cloud performance, cost efficiency, and business value. The study addresses the critical challenge enterprises face in managing cloud expenditures while maintaining operational excellence. Through comprehensive analysis of FinOps principles and practical optimization frameworks, we develop models that integrate financial accountability, technical efficiency, and business alignment. Our research demonstrates that organizations implementing structured FinOps practices achieve 25-40% cost reductions without compromising application performance. The study contributes both theoretical frameworks for understanding cloud …

Enhancing Erp Auditability And Compliance Using Permissioned Blockchain, A Framework For Transparent And Immutable Enterprise Resource Planning Systems., Manikantha Varaprasad Inakollu

Computer Science and Engineering Faculty Publications

Enterprise Resource Planning systems serve as the backbone of modern organizational operations, yet their centralized architecture creates significant challenges for auditability and regulatory compliance. This research proposes a permissioned blockchain framework to enhance ERP auditability by creating immutable, transparent, and traceable records of all system transactions and modifications. The study addresses critical gaps in current ERP systems where transaction histories can be altered, audit trails prove insufficient, and compliance verification remains cumbersome. Through examination of existing ERP limitations and blockchain capabilities, we develop an integrated architecture that maintains operational efficiency while providing cryptographic assurance of data integrity. Our framework employs …

Deep-Learning Approaches To Predict Remaining Useful Life Of Hard Disks, Rohan Mohapatra

Master's Projects

On a daily basis, data centers process huge volumes of data using inexpensive hard disks. Data stored in these disks serve a range of critical functional needs from financial, and healthcare to aerospace. As such, premature disk failure and consequent loss of data can be catastrophic. To mitigate the risk of failures, cloud storage providers perform condition-based monitoring and replace hard disks before they fail. By estimating the remaining useful life (RUL) of hard disk drives, one can predict the time-to-failure of a particular device and replace it at the right time, ensuring maximum utilization whilst reducing operational costs. We …

Echo: A Browser Extension That Runs Experimental Javascript, Prayuj Pillai

Master's Projects

Narcissus is a JavaScript interpreter written in JavaScript. While it is a good engine for experimenting with JavaScript’s design, it does not integrate easily into the browser. This project introduces ‘‘Echo’’, a browser add-on designed to execute Narcissus JavaScript files and scripts within web browsers. The project explores the performance of the Narcissus interpreter against native browser JavaScript engines and benchmarks the results, showcasing the trade-offs in running an experimental engine—the Narcissus interpreter—on the browser versus native JavaScript. Additionally, as a proof of concept, we implement taint tracking, a capability meant to boost security by preventing sensitive data from being …

A Framework For Scientific Data Indexing, Searching And Sharing, Apoorva Mohite

Master's Projects

Scientific data continues to grow. Wildfire simulation experiments performed by the WIRC team at SJSU have generated over 138 TB of data so far and it is expected to keep growing. It becomes difficult for researchers to search through that data to find the data of their interest. This data is stored on an HPC cluster that external users do not have access to. The WIRC team also conducts experiments and publishes their research, but the size of data makes it difficult to share these datasets. This project introduces a novel solution to indexing scientific data, searching through the data …

Adversarial Attacks And Defense Mechanisms In Multivariate Time-Series Forecasting For Applications In Smart And Connected Infrastructures, Pooja Krishan

Master's Projects

The advent of deep learning models has revolutionized the industry over the past decade, leading to the widespread proliferation of smart devices and infrastructures. They play a crucial role in safety-critical applications like self-driving cars and medical image analysis, sustainable technologies like power consumption prediction, and in health monitoring tools to replace industrial equipment like hard disk drives, semiconductor chips, and lithium-ion batteries. But these indispensable deep learning models can be easily fooled to give incorrect predictions with utmost conviction, leading to catastrophic failures in applications where safety is of utmost importance, and resulting in the wastage of resources in …

Fine-Tuning Large Language Models For Folder Structure Generation, Likhith Nemani

Master's Projects

Starting a new project is a significant challenge in the software development world. Building a new project skeleton and configurations will require vast amounts of time and effort. This project aims to overcome the difficulty presented by this challenge using advanced large language models, specifically fine-tuning LLMs. Our initial focus with the implementation is to use the powerful capabilities of advanced modern models to simplify and accelerate the complicated process of getting new projects started. The solution process begins with a user posting a README file to a predetermined repository. This README file then is used as a source for …

Domain Expert Bot, Amrutha Dondemadahalli Ramegowda

Master's Projects

The fast growth of artificial intelligence in human-computer interaction has been aided significantly by the introduction of conversational AI systems. This project presents a Domain Expert Bot, a multi-domain conversational bot built with advanced NLP techniques incorporated through Sentence-BERT and MapReduce to allow the bot to analyze and comprehend challenging user queries on various topics. The bot can converse on different subjects ranging from technology topics to healthcare, environment, politics, and casual discussions. It excels in understanding deep language contexts and efficiently processes large datasets, ensuring prompt and accurate responses. Furthermore, it uses advanced ranking algorithms to perform real- time …

Credit Score-Based Lending System On The Ethereum Platform, Mayuri Shimpi

Master's Projects

Traditional banking systems act as intermediaries, assessing risks and profiting from interest rate differentials. Credit scores, provided by trusted bureaus, are commonly used to evaluate the creditworthiness of borrowers. Cryptocurrencies have emerged as a significant and innovative medium due to their decentralized nature, operating without reliance on a central authority, such as a government.

This report describes a project to implement the Autonomous Lending system on the Ethereum Platform (ALOE), as proposed in [1], aiming to seamlessly integrate traditional credit scoring methodologies for evaluating a borrower's risk of default. The objective of this project report is to establish a robust …

Emulating Human Personality With Large Language Models Through Contextual Prompts And Fine-Tuning, Mrunal Zambre

Master's Projects

The quest for AI systems that can mirror the intricate aspects of human emotion and personality is crucial for enhancing their performance. This project delves into the capabilities of Large Language Models (LLMs) to mimic the Big Five personality traits in human-written essays by utilizing contextual prompts and fine-tuning methods. Diverging from traditional research in this domain, this project explores smaller, open-source LLMs, including LLaMA 2 7B chat, LLaMA 2 13B chat, and Vicuna v.15 13B, to assess their potential in personality prediction tasks, thereby making high-level personality emulation more accessible and practical for application integration. Through meticulous prompt engineering, …

Secured Data Storage Management With Deduplication In Cloud Computing, Ganesh Regoti

Master's Projects

In the cloud era, cloud storage has become a major service and the security of data and user privacy algorithms are becoming of great importance. This way, we make sure that the encrypted data is kept in the cloud storage. But, the challenges follow: First, storing encrypted data may result in ineffective utilization of cloud resources as in the provision of encrypted data, redundancy cannot be provided. Access control to the encrypted data is difficult as the underlying data is hidden and there is no metric with which the decision to share among users can be easily taken. Deduplication is …

Ranking-Based Hashtag Recommendation With Collaborative And Content-Based Filtering, Fei Pan

Master's Projects

The purpose of this project is recommending relevant hashtags for users using both Collaborative Filtering (CF) and Content-based filtering with Twitter dataset. The Twitter dataset was collected by leveraging Twitter API v2. After data preprocessing, 40,806 tweets posted by 278 users with 3,107 hashtags from 01/01/2022 to 04/30/2022 are used for model training and testing. For CF models, we will mainly focus on generating embeddings to learn about user and hashtag latent factors and finally predict a probability for unseen hashtags with most possibility will be ranked as topK items for corresponding users. In this project, Matrix Factorization (MF), Neural …

Ml-Based User Identification Through Mouse Dynamics, Rakshit Gupta

Master's Projects

User authentication and identification plays a crucial role in ensuring the security and integrity of digital systems. Traditional authentication methods, such as passwords and biometrics, have inherent limitations that can compromise system security. This research proposes a novel approach to user authentication by leveraging machine learning techniques and behavioral biometrics, specifically mouse dynamics. The primary objective is to develop a sophisticated framework that can accurately identify individuals based on their unique mouse behavior patterns. The study explores and compares multiple deep learning architectures, including Convolutional Neural Networks (CNN), Long Short-Term Memory networks (LSTM), and Transformer models, to generate embeddings from …

Analysis And Application Of Adaptive Ml Algorithms For Malware Classification, Rashmi Boddukuri

Master's Projects

Malware classification is the process of distinguishing malware samples into categories of malware families that it is associated with and remains a critical step in the process of mitigating malware-related threats. In recent years, machine learning techniques have emerged as a powerful tool for such malware classification tasks. In this study, we explore the application of adaptive machine learning models to malware classification in order to analyze and determine how they compare in performance to similar but non-adaptive algorithms. The results achieved in this study share insight into the strengths and limitations of adaptive learning models when applied towards malware …

Domain Switch On Sentiment Analysis Using Gradient Reversal Layer, Hemish Veeraboina

Master's Projects

Switching domains in sentiment analysis presents the challenge of transferring learned knowledge from one context to another without the need to label data. Traditional methods often struggle when dealing with differences in data distribution a problem known as the domain shift issue. To tackle this using Gradient Reversal Layers (GRL) has emerged as a solution for adapting to different domains in an unsupervised learning setting. This study introduces an enhancement to the standard GRL approach by incorporating a sigmoid function that gradually adjusts how intensely domain adaptation occurs during training. This upgraded GRL technique ensures controlled learning outcomes making it …

Ai-Driven Credit Card Fraud Detection With Enhanced User Interaction, Toshi Bhat

Master's Projects

This thesis describes the development and testing of a unique system for detecting credit card fraud. The system employs graph neural networks (GNNs) and a real-time user interaction platform. The primary goal of this study is to use advanced machine learning methods and interactive technologies to improve fraud detection accuracy and the speed with which users can receive assistance. GraphSAGE, a type of GNN, was trained on a simulated set of credit card transactions, allowing the system to detect and predict fraud very accurately. Simulating a real-world transaction scenario is an important aspect of the project. In this case, the …