Digital Commons Network^™

Mitigating The Shortcomings Of Language Models: Strategies For Handling Memorization & Adversarial Attacks, Aly Kassem

Electronic Theses and Dissertations

Deep learning models have recently achieved remarkable progress in Natural Language Processing (NLP), specifically in classification, question-answering, and machine translation. However, NLP models face challenges related to security and privacy. Security-wise, even small perturbations in the input can significantly impact a model's prediction. This highlights the importance of generating natural adversarial attacks to analyze the weaknesses of NLP models and bolster their robustness through adversarial training (AT). Conversely, Large Language Models (LLMs) are trained on vast amounts of data, which may include sensitive information. If exposed, this poses a risk to personal privacy. LLMs can memorize portions of their training …

Advanced Deep Learning Multivariate Multi-Time Series Framework For A Novel Covid-19 Dataset, Swastik Bagga

Electronic Theses and Dissertations

This thesis introduces an innovative framework aimed at addressing the complexities of predicting outcomes in multivariate multi time series datasets in regression analysis. By applying this framework to a novel COVID-19 dataset, it enhances predictive analytics by providing accurate forecasts for epidemic trends at regional or provincial levels, going beyond national-level analysis. The framework incorporates advanced data preprocessing, feature selection, engineering, encoding, and model architecture, effectively capturing intricate variable interactions and temporal dependencies. This makes it a powerful tool for tackling multivariate multi time series regression challenges, offering valuable insights for informed decision-making. Predicting outcomes in such datasets is challenging …

Context-Driven Behavior: Improved Contextual Reasoning For Context-Aware Agents, Christian L. Wilson

Electronic Theses and Dissertations

Over the last three decades, a considerable amount of research has been dedicated to improving an artificial agent's ability to recognize and deal effectively with context. In this paper, I discuss a framework for a novel form of contextual reasoning. Unlike existing contextual reasoning frameworks, which allow an agent to apply its contextual knowledge after it is operating in an instance of a known context, the model I discuss allows an agent to reason about context proactively. With a proactive model, an agent forecasts the future contexts it will encounter, then takes steps to ensure its behaviors are appropriate for …

Enhanced Content-Based Fake News Detection Methods With Context-Labeled News Sources, Duncan Arnfield

Electronic Theses and Dissertations

This work examined the relative effectiveness of multilayer perceptron, random forest, and multinomial naïve Bayes classifiers, trained using bag of words and term frequency-inverse dense frequency transformations of documents in the Fake News Corpus and Fake and Real News Dataset. The goal of this work was to help meet the formidable challenges posed by proliferation of fake news to society, including the erosion of public trust, disruption of social harmony, and endangerment of lives. This training included the use of context-categorized fake news in an effort to enhance the tools’ effectiveness. It was found that term frequency-inverse dense frequency provided …

A Bridge Between Graph Neural Networks And Transformers: Positional Encodings As Node Embeddings, Bright Kwaku Manu

Electronic Theses and Dissertations

Graph Neural Networks and Transformers are very powerful frameworks for learning machine learning tasks. While they were evolved separately in diverse fields, current research has revealed some similarities and links between them. This work focuses on bridging the gap between GNNs and Transformers by offering a uniform framework that highlights their similarities and distinctions. We perform positional encodings and identify key properties that make the positional encodings node embeddings. We found that the properties of expressiveness, efficiency and interpretability were achieved in the process. We saw that it is possible to use positional encodings as node embeddings, which can be …

Geo-Location Informed Team Formation Using Gnn, Karan Saxena

Electronic Theses and Dissertations

Establishing a competent team is crucial to the success of a project and is influenced by skill distribution and geographic proximity. A team not only benefits from the shared knowledge amongst the team members derived from geographic closeness but also affects the outcome of the project the team is assigned to perform. A team benefits by sharing resources among each member, collaborating efficiently on a given task, brainstorming on an idea more effectively and saving time and money for both the team members and the organization. This thesis uses a neural-based multi-label classifier after a spatial team formation that uses …

Enhancing Search Engine Results: A Comparative Study Of Graph And Timeline Visualizations For Semantic And Temporal Relationship Discovery, Muhammad Shahiq Qureshi

Electronic Theses and Dissertations

In today’s digital age, search engines have become indispensable tools for finding information among the corpus of billions of webpages. The standard that most search engines follow is to display search results in a list-based format arranged according to a ranking algorithm. Although this format is good for presenting the most relevant results to users, it fails to represent the underlying relations between different results. These relations, among others, can generally be of either a temporal or semantic nature. A user who wants to explore the results that are connected by those relations would have to make a manual effort …

Matches Made In Heaven Or Somewhere: Personalized Query Refinement Gold Standard Generation Using Transformers, Yogeswar Lakshmi Narayanan

Electronic Theses and Dissertations

The foremost means of information retrieval, search engines, have difficulty searching into knowledge repositories, e.g., the web, because they are not tailored to the users' differing information needs. User queries are, more often than not, under-specified or contain ambiguous terms that also retrieve irrelevant documents. Query refinement is the process of transforming users' queries into new refined versions without semantic drift to enhance the relevance of search results. Prior query refiners have been benchmarked on ad-hoc web retrieval datasets following weak assumptions that users' input queries improve gradually within a search session. Existing methods also have employed additional metadata, such …

Many-To-One: Transformer-Based Unsupervised Anomaly Detection And Localization On Industrial Images, Naga Jyothirmayee Dodda

Electronic Theses and Dissertations

Anomaly detection is of utmost importance in the realm of industrial defect identification, particularly when employing computer vision-based inspection mechanisms within quality control systems. This research introduces the Many-to-One (M2O) framework, which relies on a multi-level transformer encoder combined with a single transformer decoder, which forms many-to-one relation in the framework for detecting and localizing anomalies. The rise of Industry 4.0 and electric vehicles has increased interest in this area. Although previous research has made significant contributions, challenges still exist in this area. It is crucial to develop models that can generalize well and overcome time complexity problems that affect …

Anomaly Detection In Large Datasets: A Case Study In Loan Defaults, Rayhaan Pirani

Electronic Theses and Dissertations

Given the rise in loan defaults, especially after the COVID-19 pandemic, it is necessary to predict if customers might default on a loan for risk management. This thesis proposes an early warning system architecture using anomaly detection based on the unbalanced nature of loan default data in the real world. Most customers do not default on their loans; only a tiny percentage do, resulting in an unbalanced dataset. We aim to evaluate potential anomaly detection methods for their suitability in handling unbalanced datasets. We conduct a comparative study on different anomaly detection approaches on four balanced and unbalanced datasets. We …

An Approach Of Lip Synchronization With Facial Expression Rendering For An Eca, Mohammed Maaz Mohammed Shoieb Khan

Electronic Theses and Dissertations

E-Tutoring systems have emerged as effective tools for remote learning and personalized education. However, to foster engaging and interactive experiences, there is a need to enhance the communication and expressiveness of virtual tutors within these systems. This research focuses on integrating facial expression animation and lip-syncing capabilities into an e-tutoring system, aiming to improve the realism and effectiveness of virtual tutor interactions. The study presents a novel approach to animating the facial expressions of the virtual tutor’s avatar using computer graphics, 3D animation, and computer vision techniques. Working with 3D files and using them to interpolate each face vertex are …

Transmission Power Based Congestion Control Using Q-Learning Algorithm In Vehicular Ad Hoc Networks (Vanet), Pooja Chandrasekharan

Electronic Theses and Dissertations

To enhance road safety, Vehicular ad hoc networks (VANETs), an emerging wireless technology used for vehicle-to-vehicle and vehicle-to-infrastructure communication, are essential components to reduce road accidents and traffic congestion in Intelligent Transportation Systems (ITS). It also provides additional services to vehicles and their users. However, vehicles must balance awareness and congestion control in a dynamic environment to efficiently transmit basic safety messages (BSMs) and event-driven warnings. The limited channel capacity makes the reliable delivery of BSMs a challenging problem for VANETs. This paper aims to optimize the performance of VANETs by effectively managing channel load and reducing congestion by maintaining …

An End-To-End Hybrid Approach To Automatic And Semi-Automatic Image Annotation, Roisul Islam Rumi

Electronic Theses and Dissertations

This thesis presents a new end-to-end hybrid machine learning (ML) and deep learning (DL) approach for semi-automatic image annotation (SAIA) and automatic image annotation (AIA) in industrial assembly line setups. Image annotation refers to adding descriptive labels or tags to an image to provide information about the objects and features present in the image. On a high level, the proposed system uses the following steps to annotate images. The first step involves using an ML algorithm, Haar cascade, to split an image into smaller regions of interest (ROI) based on the object of interest, in our case, the connectors of …

Accessible Autonomy: Exploring Inclusive Autonomous Vehicle Design And Interaction For People Who Are Blind And Visually Impaired, Paul D. S. Fink

Electronic Theses and Dissertations

Autonomous vehicles are poised to revolutionize independent travel for millions of people experiencing transportation-limiting visual impairments worldwide. However, the current trajectory of automotive technology is rife with roadblocks to accessible interaction and inclusion for this demographic. Inaccessible (visually dependent) interfaces and lack of information access throughout the trip are surmountable, yet nevertheless critical barriers to this potentially lifechanging technology. To address these challenges, the programmatic dissertation research presented here includes ten studies, three published papers, and three submitted papers in high impact outlets that together address accessibility across the complete trip of transportation. The first paper began with a thorough …

A Quantitative Visualization Tool For The Assessment Of Mammographic Risky Dense Tissue Types, Margaret R. Mccarthy

Electronic Theses and Dissertations

Breast cancer is the second most occurring cancer type and is ranked fifth in terms of mortality. X-ray mammography is the most common methodology of breast imaging and can show radiographic signs of cancer, such as masses and calcifcations. From these mammograms, radiologists can also assess breast density, which is a known cancer risk factor. However, since not all dense tissue is cancer-prone, we hypothesize that dense tissue can be segregated into healthy vs. risky subtypes. We propose that risky dense tissue is associated with tissue microenvironment disorganization, which can be quantified via a computational characterization of the whole breast …

Proposing A Measure Of Ethicality For Humans And Ai, Alejandro Jorge Napolitano Jawerbaum

Electronic Theses and Dissertations

Smarter people or intelligent machines are able to make more accurate inferences about their environment and other agents more efficiently than less intelligent agents. Formally: ‘Intelligence measures an agent’s ability to achieve goals in a wide range of environments.’ (Legg, 2008)

In this dissertation we extend this definition to include ethical behaviour and we will offer a mathematical formalism and a way to estimate how ethical an action is or will be, both for a human and for a computer, by calculating the expected values of random variables. Formally, we propose the following measure of ethicality, which is computable, or …

Controllable Language Generation Using Deep Learning, Rohola Zandie

Electronic Theses and Dissertations

The advent of deep neural networks has sparked a revolution in Artificial Intelligence (AI), notably with the creation of Transformer models like GPT-X and ChatGPT. These models have surpassed previous methods in various Natural Language Processing (NLP) tasks. As the NLP field evolves, there is a need to further understand and question the capabilities of these models. Text generation, a crucial part of NLP, remains an area where our comprehension is limited while being critical in research.

This dissertation focuses on the challenging problem of controlling the general behaviors of language models such as sentiment, topical focus, and logical reasoning. …

A Data-Driven Multi-Regime Approach For Predicting Real-Time Energy Consumption Of Industrial Machines., Abdulgani Kahraman

Electronic Theses and Dissertations

This thesis focuses on methods for improving energy consumption prediction performance in complex industrial machines. Working with real-world industrial machines brings several challenges, including data access, algorithmic bias, data privacy, and the interpretation of machine learning algorithms. To effectively manage energy consumption in the industrial sector, it is essential to develop a framework that enhances prediction performance, reduces energy costs, and mitigates air pollution in heavy industrial machine operations. This study aims to assist managers in making informed decisions and driving the transition towards green manufacturing. The energy consumption of industrial machinery is substantial, and the recent increase in CO2 …

Cannabidiol Tweet Miner: A Framework For Identifying Misinformation In Cbd Tweets., Jason Turner

Electronic Theses and Dissertations

As regulations surrounding cannabis continue to develop, the demand for cannabis-based products is on the rise. Despite not producing the psychoactive effects commonly associated with THC, products containing cannabidiol (CBD) have gained immense popularity in recent years as a potential treatment option for a range of conditions, particularly those associated with pain or sleep disorders. However, due to current federal policies, these products have yet to undergo comprehensive safety and efficacy testing. Fortunately, utilizing advanced natural language processing (NLP) techniques, data harvested from social networks have been employed to investigate various social trends within healthcare, such as disease tracking and …

Evaluating Chatgpt For Recommendation: How Does The Ability To Converse Impact Recommendation?, Kyle Spurlock

Electronic Theses and Dissertations

Recommendation algorithms have become an absolute necessity in the modern world to avoid information overload. However, the interaction between the human and the system is largely superficial and without any real contact. If you are given poor recommendations, you have no choice but to sift through mountains of content on your own until the model learns to accommodate your tastes more. This is bad for business as well as the consumer. Recently, large language models like ChatGPT have seen a significant rise in popularity due to their ease of use and wide range of knowledge. It has now become nearly …

Extending The Work Of Dt-Fixup: Examining The Effects Of Powernorm And Madgrad Optimization On Dt-Fixup Performance, Prem Shanker Mohan

Electronic Theses and Dissertations

With the introduction of the attention technique, the Bidirectional Encoder Representations from Transformers (BERT) have greatly advanced the study of solving sequence-to-sequence tasks in Natural Language Processing (NLP). When the task-specific annotations are limited, the NLP tasks are commonly performed by pre-training a model using the transformer technique on large-scale general corpora, followed by fine-tuning the model on domain-specific data. Instead of using shallow neural components for fine-tuning, additional transformer layers could be introduced into the architecture. Recent research shows that, by resolving some initialization and optimization issues, these augmented transformer layers could lead to performance gains despite of the …

Reinforcement Learning-Based Data Rate Congestion Control For Vehicular Ad-Hoc Networks, Gnana Shilpa Nuthalapati

Electronic Theses and Dissertations

Vehicular Ad-Hoc Network(VANET) is an emerging wireless technology vital to the Intelligent Transportation System(ITS) for vehicle-to-vehicle and vehicle-to-infrastructure communication. An ITS is an advanced solution that aims to deliver innovative services pertaining to various transportation modes and traffic management. Its objective is to enhance user awareness, promote safety, and enable more efficient and coordinated utilization of transport networks. ITS aims to mitigate traffic problems and improve the safety of transport by preventing unexpected events. When the vehicle density, i.e., the number of vehicles communicating in a wireless channel, increases, the channel faces congestion resulting in unreliable safety applications. Various decentralized …

Classifying Galaxy Images Using Improved Residual Networks, Jaykumar Patel

Electronic Theses and Dissertations

The field of astronomy has made tremendous progress in recent years thanks to advancements in technology and the development of sophisticated algorithms. One area of interest for astronomers is the classification of galaxy morphology, which involves categorizing galaxies based on their visual appearance. However, with the sheer number of galaxy images available, it would be a daunting task to manually classify them all. To address this challenge, a novel Residual Neural Network (ResNet) model, called ResNet Var, that can automatically classify galaxy images is proposed in this study. Galaxy Zoo 2 dataset is used in this research, which contains over …

Cross-Blockchain Technology For An Interoperable And Scalable Digital Contact Tracing, Farbod Behnaminia

Electronic Theses and Dissertations

COVID-19 pandemic has highlighted the importance of contact tracing as a tool for controlling the spread of the virus, but it has also raised concerns about the privacy and security of personal information. Blockchain technology, with its immutability and security features, has the potential to address these concerns. However, traditional blockchain solutions may not be sufficient to protect sensitive personal information, especially when it comes to interoperability with other chains that may have different privacy standards. Cross-blockchain technology, such as the interoperability feature of the Polkadot network, allows for the creation of a decentralized and distributed contact tracing system that …

Personalized Eca Tutoring With Self-Adjusted Pomdp Policies And User Clustering, Ashwitha Vichuly Jawahar

Electronic Theses and Dissertations

An Embodied Conversational Agent (ECA) is an intelligent agent that enables realtime human/computer interaction in natural language. For its rich style of communication, ECA is particularly popular and useful in applications such as education, e-commerce, healthcare, finance, marketing, and business, where a human-like conversation is more attractive to users than traditional keyboard-based interaction. The interest in using ECA in e-learning has become even stronger since the COVID-19 outbreak, and a preliminary investigation has been started by our research group to extend collaborative learning in a virtual environment with personalized ECA tutoring. This thesis document first highlights the prior work of …

A Decentralized Data Evaluation Technique In Federated Learning, Laveen Bhatia

Electronic Theses and Dissertations

Deep Learning is one of the most revolutionary concepts in the field of Artificial Intelligence, allowing us to train a Machine Learning model for almost any type of problem using any type of data. Federated Learning (FL) is a type of distributed Deep Learning framework in which the model is trained locally on each device, and only the trained gradients, also known as “local updates”, are sent to a central server that aggregates them and creates a global model. This helps in preserving the data privacy of the user as the local data never leaves the local device. It has …

An Investigation Into Machine Learning Techniques For Designing Dynamic Difficulty Agents In Real-Time Games, Ryan Adare Dunagan

Electronic Theses and Dissertations

Video games are an incredibly popular pastime enjoyed by people of all ages world wide. Many different kinds of games exist, but most games feature some elements of the player overcoming some challenge, usually through gameplay. These challenges are insurmountable for some people and may turn them off to video games as a pastime. Games can be made more accessible to players of little skill and/or experience through the use of Dynamic Difficulty Adjustment (DDA) systems that adjust the difficulty of the game in response to the player’s performance. This research seeks to establish the effectiveness of machine learning techniques …

Patient Movement Monitoring Based On Imu And Deep Learning, Mohsen Sharifi Renani

Electronic Theses and Dissertations

Osteoarthritis (OA) is the leading cause of disability among the aging population in the United States and is frequently treated by replacing deteriorated joints with metal and plastic components. Developing better quantitative measures of movement quality to track patients longitudinally in their own homes would enable personalized treatment plans and hasten the advancement of promising new interventions. Wearable sensors and machine learning used to quantify patient movement could revolutionize the diagnosis and treatment of movement disorders. The purpose of this dissertation was to overcome technical challenges associated with the use of wearable sensors, specifically Inertial Measurement Units (IMUs), as a …

The Minimum Consistent Spanning Subset Problem On Trees, Parham Khamsepour

Electronic Theses and Dissertations

Given a vertex-colored edge-weighted graph, the minimum consistent subset (MCS) problem asks for a minimum subset S of vertices such that every vertex v not in S has the same color as its nearest neighbor in S. This problem has applications in clustering and classification algorithms, specially in finding the optimal number of clusters in k-clustering algorithms. This problem is NP-complete. A recent result of Dey, Maheshwari, and Nandy (2021) gives a polynomial-time algorithm for the MCS problem on trees. In thesis we study the MCS problem on different settings, and discuss some of the shortcomings of the MCS problem …

Approximating Average Bounded-Angle Minimum Spanning Trees, Patrick Stephen Devaney

Electronic Theses and Dissertations

Motivated by the problem of orienting directional antennas in wireless communication networks, we study average bounded-angle minimum spanning trees. Let P be a set of points in the plane and let α be an angle. An α-spanning tree (α-ST) of P is a spanning tree of the complete Euclidean graph induced by P with the restriction that all edges incident to each point p in P lie in a wedge of angle α with apex p. An α-minimum spanning tree (α-MST) of P is an α-ST with minimum total edge length. An average-α-spanning tree (denoted by avg-α-ST) is a spanning …