Computer Sciences Commons^™

Efficiently Learning An Encoder That Classifies Token Replacements And Masked Permuted Network-Based Bigru Attention Classifier For Enhancing Sentiment Classification Of Scientific Text, Muhammad Inaam Ul Haq, Khalid Mahmood, Qianmu Li, Ashok Kumar Das, Sachin Shetty, Majid Hussain

VMASC Publications

The exponential growth of scientific literature in digital repositories poses challenges in interpreting complex attitudes within academic texts. Traditional sentiment analysis methods often struggle with nuanced word meanings due to contextual variations. To address this, we propose the Electra-MPNet-based BiGRU attention classifier that extracts the high-level semantic features from citation sentences using the combined strength of Electra and MPNet encoder layers. These features are then combined to extract long-range dependencies through a stacked BiGRU layer. A linear attention mechanism is imposed to estimate the attention weights and context vector which enables the model to selectively focus on relevant information. The …

Preparing Healthcare Education For An Ai-Augmented Future, Jiajie Zhang, Susan H Fenton

Faculty, Staff and Student Publications

Artificial intelligence (AI) fundamentally transforms healthcare education as a knowledge enterprise, creating a distributed cognitive system composed of the human brain, which remains relatively unchanged, and AI-based knowledge and cognitive functions, which have accelerated exponentially in scale and power. Education must focus on developing skills to collaborate with AI and on achieving outcomes like problems solved and discoveries made. Curriculum and education policies also need to adapt to this transformation.

Pneumothorax Detection And Segmentation From Chest X-Ray Radiographs Using A Patch-Based Fully Convolutional Encoder-Decoder Network, Jakov Ivan S. Dumbrique, Reynan Hernandez, Juan Miguel L. Cruz, Ryan M. Pagdanganan, Prospero C. Naval

Mathematics Faculty Publications

Pneumothorax, a life-threatening condition characterized by air accumulation in the pleural cavity, requires early and accurate detection for optimal patient outcomes. Chest X-ray radiographs are a common diagnostic tool due to their speed and affordability. However, detecting pneumothorax can be challenging for radiologists because the sole visual indicator is often a thin displaced pleural line. This research explores deep learning techniques to automate and improve the detection and segmentation of pneumothorax from chest X-ray radiographs. We propose a novel architecture that combines the advantages of fully convolutional neural networks (FCNNs) and Vision Transformers (ViTs) while using only convolutional modules to …

Federated Learning: Overview, Strategies, Applications, Tools And Future Directions, Betul Yurdem, Murat Kuzlu, Mehmet Kemal Gullu, Maliha Tabassum

Engineering Technology Faculty Publications

Federated learning (FL) is a distributed machine learning process, which allows multiple nodes to work together to train a shared model without exchanging raw data. It offers several key advantages, such as data privacy, security, efficiency, and scalability, by keeping data local and only exchanging model updates through the communication network. This review paper provides a comprehensive overview of federated learning, including its principles, strategies, applications, and tools along with opportunities, challenges, and future research directions. The findings of this paper emphasize that federated learning strategies can significantly help overcome privacy and confidentiality concerns, particularly for high-risk applications.

A Benchmark Framework For Data Visualization And Explainable Ai (Xai), Murat Kuzlu, Gokcen Ozdemir, Umut Ozdemir

Engineering Technology Faculty Publications

This research introduces a benchmark framework, called EDUMX, designed for machine learning (ML)-based forecasting and XAI tasks, leveraging the Streamlit open-source Python library. The framework offers a comprehensive suite of functionalities, including data loading, feature selection, relationship analysis, data preprocessing, model selection, metric evaluation, training, and real-time monitoring. Users can easily upload data in diverse formats, explore relationships between variables, preprocess data using various techniques, and assess the performance of the ML model using customizable metrics. With its user-friendly interface, this framework offers invaluable insights for forecasting tasks in various domains, catering to the evolving needs of predictive analytics. EDUMX …

Towards Algorithmic Justice: Human Centered Approaches To Artificial Intelligence Design To Support Fairness And Mitigate Bias In The Financial Services Sector, Jihyun Kim

CMC Senior Theses

Artificial Intelligence (AI) has positively transformed the Financial services sector but also introduced AI biases against protected groups, amplifying existing prejudices against marginalized communities. The financial decisions made by biased algorithms could cause life-changing ramifications in applications such as lending and credit scoring. Human Centered AI (HCAI) is an emerging concept where AI systems seek to augment, not replace human abilities while preserving human control to ensure transparency, equity and privacy. The evolving field of HCAI shares a common ground with and can be enhanced by the Human Centered Design principles in that they both put humans, the user, at …

Road Extraction On Remote Sensing Imagery: Historical Mapping Of The Brazilian Amazon, Jonas Paiva Botelho Jr

Graduate Theses/Dissertations

This work proposes an artificial intelligence model based on U-Net architecture to map road networks in the Brazilian Amazon. Over the years, the Amazon region has been heavily exploited, leading to increased deforestation rates, contributing to CO2 emissions, amplifying global warming, and causing a disturbance in local fauna and flora. The expansion into the forest by illegal miners, loggers, and land grabbers can be tracked down by the construction of roads, which we can refer to as the arteries of deforestation. Previous works on the matter proposed algorithms that use high-resolution imagery to map roads precisely. However, this work approach …

Optimizing Sports Outcome Prediction Through Feature Engineering And Machine Learning, Vitor S. Freitas

Graduate Theses/Dissertations

The challenge of predicting the outcome of a team game lies in the high complexity and dynamics of the sports data. This thesis focuses on the aspect of using feature engineering and the genetic algorithm to predict the winner and the score of various sports events. Generally, it deals with how machine learning algorithms are combined with state-of-the-art feature engineering techniques in sports datasets derived from various sports disciplines. In this thesis, five different machine learning models have been applied, classification and regression trees (CART), random forest (RF), stochastic gradient boosting (SGB), eXtreme gradient boosting (XGBoost), and extreme learning machine …

A Benchmark Framework For Data Visualization And Explainable Ai (Xai), Murat Kuzlu, Gokcen Ozdemir, Umut Ozdemir

Engineering Technology Faculty Publications

This research introduces a benchmark framework, called EDUMX, designed for machine learning (ML)- based forecasting and XAI tasks, leveraging the Streamlit open-source Python library. The framework offers a comprehensive suite of functionalities, including data loading, feature selection, relationship analysis, data preprocessing, model selection, metric evaluation, training, and real-time monitoring. Users can easily upload data in diverse formats, explore relationships between variables, preprocess data using various techniques, and assess the performance of the ML model using customizable metrics. With its user-friendly interface, this framework offers invaluable insights for forecasting tasks in various domains, catering to the evolving needs of predictive analytics. …

Identifying And Predicting Patterns Of Snowpack Ripening With Machine Learning Methods, Clement Cherblanc

Graduate Student Theses, Dissertations, & Professional Papers

The timing of water release from the snowpack plays key roles in ecosystem services, groundwater recharge, and water resource management. However, two internal barriers in a standing snowpack must be overcome before runoff can outflow from the base: 1) the cold content must be exhausted, and 2) the interconnected network of snow grains must be filled with liquid water to residual saturation. Expressing the liquid water as latent heat allows the two barriers to be grouped as an energy (J/m²) to define a snowpack’s Runoff Energy Hurdle (REH). The growth and loss of REH is driven by evolution of pore …

Advanced Techniques In Time Series Forecasting: From Deterministic Models To Deep Learning, Xue Bai

Graduate Theses, Dissertations, and Problem Reports (ETD)

This dissertation discusses three instances of temporal prediction, applied to population dynamics and deep learning.

In population modeling, dynamic processes are frequently represented by systems of differential equations, allowing for the analysis of various phenomena. The first application explores modeling cloned hematopoiesis in chronic myeloid leukemia (CML) via a nonlinear system of differential equations. By tracking the evolution of different cell compartments, including cycling and quiescent stem cells, progenitor cells, differentiated cells, and terminally differentiated cells, the model captures the transition from normal hematopoiesis to the chronic and accelerated-acute phases of CML. Three distinct non-zero steady states are identified, representing …

An Enhanced Deep Autoencoder For Flight Delay Prediction, Desmond B. Bisandu, Dan Andrei Soviani-Sitoiu, Irene Moulitsas

Journal of Aviation/Aerospace Education & Research

Accurate and timely flight delay prediction cannot be overemphasized because of the ever-increasing demand for air travel and its importance in deploying intelligent transportation systems. Nonetheless, there has not been a universal solution to the problem, as more intelligent flight decision systems are required for the aviation industry's future growth. Existing flight delay classification and prediction approaches are mainly shallow traffic models and do not satisfy many applications in the real world. Our motivation to rethink the deep architecture model for predicting flight delays emanates from the problem. In this research, we proposed a technique that modified stacked autoencoder architecture …

Deapsecure Computational Training For Cybersecurity: Progress Toward Widespread Community Adoption, Wirawan Purwanto, Bahador Dodge, Karina Arcaute, Masha Sosonkina, Hongyi Wu

Electrical & Computer Engineering Faculty Publications

The Data-Enabled Advanced Computational Training Program for Cybersecurity Research and Education (DeapSECURE) is a non-degree training consisting of six modules covering a broad range of cyberinfrastructure techniques, including high performance computing, big data, machine learning and advanced cryptography, aimed at reducing the gap between current cybersecurity curricula and requirements needed for advanced research and industrial projects. Since 2020, these lesson modules have been updated and retooled to suit fully-online delivery. Hands-on activities were reformatted to accommodate self-paced learning. In this paper, we summarize the four years of the project comparing in-person and on-line only instruction methods as well as outlining …

Beyond Binary: Revealing Variations In Islamophobic Content With Hierarchical Multi-Class Classification, Esraa Aldreabi, Khawlah M. Harahsheh, Mukul Dev Chhangani, Chung-Hao Chen, Jeremy Blackburn

Electrical & Computer Engineering Faculty Publications

In the digital age, the rise of Islamophobia-marked by an irrational fear or discrimination against Islam and Muslims-has emerged as a pressing issue, especially on social media platforms. In this paper we employs a multi-class classification system, moving beyond traditional binary models. We categorize Islamophobic content into three main classes and various subclasses, covering a range from subtle biases to explicit incitement. Comparative analysis of data from Reddit and Twitter illuminates the distinct prevalence and types of Islamophobic content specific to each platform. This paper deepens our understanding of digital Islamophobia and provides insights for crafting targeted online counter strategies. …

Combination Chemotherapy Optimization With Discrete Dosing, Temitayo Ajayi, Seyedmohammadhossein Hosseinian, Andrew J Schaefer, Clifton D Fuller

Faculty, Staff and Student Publications

Chemotherapy drug administration is a complex problem that often requires expensive clinical trials to evaluate potential regimens; one way to alleviate this burden and better inform future trials is to build reliable models for drug administration. This paper presents a mixed-integer program for combination chemotherapy (utilization of multiple drugs) optimization that incorporates various important operational constraints and, besides dose and concentration limits, controls treatment toxicity based on its effect on the count of white blood cells. To address the uncertainty of tumor heterogeneity, we also propose chance constraints that guarantee reaching an operable tumor size with a high probability in …

Runtime Support For Cpu-Gpu High-Performance Computing On Distributed Memory Platforms, Polykarpos Thomadakis, Nikos Chrisochoides

Computer Science Faculty Publications

Hardware heterogeneity is here to stay for high-performance computing. Large-scale systems are currently equipped with multiple GPU accelerators per compute node and are expected to incorporate more specialized hardware. This shift in the computing ecosystem offers many opportunities for performance improvement; however, it also increases the complexity of programming for such architectures. This work introduces a runtime framework that enables effortless programming for heterogeneous systems while efficiently utilizing hardware resources. The framework is integrated within a distributed and scalable runtime system to facilitate performance portability across heterogeneous nodes. Along with the design, this paper describes the implementation and optimizations performed, …

Bayesian Neural Netwok Variational Autoencoder Inverse Mapper (Bnn-Vaim) And Its Application In Compton Form Factors Extraction, Md Fayaz Bin Hossen, Tareq Alghamdi, Manal Almaeen, Yaohang Li

Computer Science Faculty Publications

We extend the Variational Autoencoder Inverse Mapper (VAIM) framework for the inverse problem of extracting Compton Form Factors (CFFs) from deeply virtual exclusive reactions, such as the unpolarized Deeply virtual exclusive scattering (DVCS) cross section. VAIM is an end-to-end deep learning framework to address the solution ambiguity issue in ill-posed inverse problems, which comprises of a forward mapper and a backward mapper to simulate the forward and inverse processes, respectively. In particular, we incorporate Bayesian Neural Network (BNN) into the VAIM architecture (BNN-VAIM) for uncertainty quantification. By sampling the weights and biases distributions of the BNN in the backward mapper …

Understaning Low Vision Graphical Perception Of Bar Charts, Yash Prakash, Akshay Kolgar Nayak, Sampath Jayarathna, Hae-Na Lee, Vikas Ashok

Computer Science Faculty Publications

Bar charts are widely used for their simplicity in data representation, prompting numerous studies to explore and model how users interact with and perceive bar chart information. However, these studies have predominantly focused on sighted users, with a few also targeting blind screen-reader users, whereas the graphical perception of low-vision screen magnifier users is still an uncharted research territory. We fill this knowledge gap in this paper by designing four experiments for a laboratory study with 25 low-vision participants to examine their graphical perception while interacting with bar charts. For our investigation, we built a custom screen magnifier-based logger that …

Improving Usability Of Data Charts In Multimodal Documents For Low Vision Users, Yash Prakash, Akshay Kolgar Nayak, Shoaib Mohammed Alyaan, Pathan Aseef Khan, Hae-Na Lee, Vikas Ashok

Computer Science Faculty Publications

Data chart visualizations and text are often paired in news articles, online blogs, and academic publications to present complex data. While chart visualizations offer graphical summaries of the data, the accompanying text provides essential context and explanation. Associating information from text and charts is straightforward for sighted users but presents significant challenges for individuals with low vision, especially on small-screen devices such as smartphones. The visual nature of charts coupled with the layout of the text inherently makes it difficult for low vision users to mentally associate chart data with text and comprehend the content due to their dependence on …

Exacfs - A Cil Method To Mitigate Catastrophic Forgetting, S. Balasubramanian, Sai Subramaniam M., Sai Sriram Talasu, Manepalli Pranav Phanindra Sai, Yedu P. Krishna, Darshan Gera, Ravi Mukkamala

Computer Science Faculty Publications

Deep neural networks (DNNs) excel at learning from static datasets but struggle with continual learning, where data arrives sequentially. Catastrophic forgetting, the phenomenon of forgetting previously learned knowledge, is a primary challenge. This paper introduces EXponentially Averaged Class-wise Feature Significance (EXACFS) to mitigate this issue in the class incremental learning (CIL) setting. By estimating the significance of model features for each learned class using loss gradients, gradually aging the significance through the incremental tasks and preserving the significant features through a distillation loss, EXACFS effectively balances remembering old knowledge (stability) and learning new knowledge (plasticity). Extensive experiments on CIFAR-100 and …

Learning Optimal Inter-Class Margin Adaptively For Few-Shot Class-Incremental Learning Via Neural Collapse-Based Meta-Learning, Hang Ran, Weijun Li, Lusi Li, Songsong Tian, Xin Ning, Prayag Tiwari

Computer Science Faculty Publications

Few-Shot Class-Incremental Learning (FSCIL) aims to learn new classes incrementally with a limited number of samples per class. It faces issues of forgetting previously learned classes and overfitting on few-shot classes. An efficient strategy is to learn features that are discriminative in both base and incremental sessions. Current methods improve discriminability by manually designing inter-class margins based on empirical observations, which can be suboptimal. The emerging Neural Collapse (NC) theory provides a theoretically optimal inter-class margin for classification, serving as a basis for adaptively computing the margin. Yet, it is designed for closed, balanced data, not for sequential or few-shot …

Osfs-Vague: Online Streaming Feature Selection Algorithm Based On A Vague Set, Jie Yang, Zhijun Wang, Guoyin Wang, Yanmin Liu, Yi He, Di Wu

Computer Science Faculty Publications

Online streaming feature selection (OSFS), as an online learning manner to handle streaming features, is critical in addressing high-dimensional data. In real big data-related applications, the patterns and distributions of streaming features constantly change over time due to dynamic data generation environments. However, existing OSFS methods rely on presented and fixed hyperparameters, which undoubtedly lead to poor selection performance when encountering dynamic features. To make up for the existing shortcomings, the authors propose a novel OSFS algorithm based on vague set, named OSFS-Vague. Its main idea is to combine uncertainty and three-way decision theories to improve feature selection from the …

Sparse Representation Learning For Temporal Networks, Maxwell Mcneil

Electronic Theses & Dissertations (2024 - present)

Temporal networks arise in many domains including activity of social network users, sensor network readings over time, and time course gene expression within the interaction network of a model organism. Data of this type contains a wealth of prior information such as the connectivity among nodes (e.g., a friendship graph), and prior knowledge of expected temporal patterns (e.g., periodicity). Modeling these temporal and network patterns jointly is essential for state-of-the-art performance in temporal network data analysis and mining. Sparse dictionary encoding is one modeling approach for such underlying patterns. However, most classical approaches consider only one dimension of the data …

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 …

Reducing Food Scarcity: The Benefits Of Urban Farming, S.A. Claudell, Emilio Mejia

Journal of Nonprofit Innovation

Urban farming can enhance the lives of communities and help reduce food scarcity. This paper presents a conceptual prototype of an efficient urban farming community that can be scaled for a single apartment building or an entire community across all global geoeconomics regions, including densely populated cities and rural, developing towns and communities. When deployed in coordination with smart crop choices, local farm support, and efficient transportation then the result isn’t just sustainability, but also increasing fresh produce accessibility, optimizing nutritional value, eliminating the use of ‘forever chemicals’, reducing transportation costs, and fostering global environmental benefits.

Imagine Doris, who is …

Lesion Detection In Women Breast’S Dynamic Contrast-Enhanced Magnetic Resonance Imaging Using Deep Learning, Sudarshan Saikia, Tapas Si, Darpan Deb, Kangkana Bora, Saurav Mallik, Ujjwal Maulik, Zhongming Zhao

Faculty, Staff and Student Publications

Breast cancer is one of the most common cancers in women and the second foremost cause of cancer death in women after lung cancer. Recent technological advances in breast cancer treatment offer hope to millions of women in the world. Segmentation of the breast's Dynamic Contrast-Enhanced Magnetic Resonance Imaging (DCE-MRI) is one of the necessary tasks in the diagnosis and detection of breast cancer. Currently, a popular deep learning model, U-Net is extensively used in biomedical image segmentation. This article aims to advance the state of the art and conduct a more in-depth analysis with a focus on the use …

Tiny Machine Learning For Underwater Image Enhancement: Pruning And Quantizaition Approach, Dr Khaled Nagaty, The British University In Egypt, Andreas Pester Dr

Computer Science

Many people have expressed an interest in underwater image processing in a variety of fields, including underwater vehicle control, archaeology, marine biological studies, etc. Underwater exploration is becoming an increasingly important element of our lives, with applications ranging from underwater marine and creature research to pipeline and communication logistics, military use, touristic and entertainment use. Underwater images suffer from poor visibility, distortion, and poor quality for a variety of causes, including light propagation. The major issue arises when these images must be captured at depths greater than 500 feet and artificial lighting needs to be provided. Efficient algorithms and models …

Deep Learning For Microbiome-Based Integrative Modeling And Microbial Biomarkers Identification, Sen Yang

Statistical Science Theses and Dissertations

The human microbiome, comprising trillions of microorganisms, plays a pivotal role in modulating host physiology via molecular and metabolite exchanges. One of the major challenges in this field lies in the effective integration of microbiome and metabolomics data, an achievement that holds the promise of substantially enhancing the precision of disease prediction. However, many datasets prioritize microbiome data while neglecting paired metabolome information. Additionally, the prevalent analytical tools face challenges in effectively merging these intricate datasets, leading to possible misinterpretations and reduced prediction accuracies.

To address these challenges, the first part of this research introduces the Microbiome-based Supervised Contrastive Learning …

Roadside Lidar Data Processing For Intelligent Transportation System, Md Parvez Mollah

Computer Science ETDs

Roadside LiDAR (Light Detection and Ranging) sensors are recently being explored for Intelligent Transportation System aiming at safer and faster traffic management and vehicular operations. However, massive data volume, occlusion, and limited viewing angles are significant obstacles to the widespread use of roadside LiDARs. In this dissertation, we address three major challenges to enable applications of Intelligent Transportation System through roadside LiDAR data: (i) real-time transmission of the massive point-cloud data from the roadside LiDAR devices to the cloud using 5G network, (ii) mitigating sensor occlusion problem to increase coverage and detect events occurred in occluded regions of a sensor, …

An Empirical Study Of Machine Learning Techniques For Accurate Stock Price Forecasting, Daniel Paliulis, Hari Patchigolla

Honors Scholar Theses

This paper presents a comprehensive approach to predicting future stock prices of companies using machine learning and time series analysis. The research problem is centered around addressing the complexity and emotion-driven nature of stock investment decisions. To create an objective determinant in stock decisions, we propose a machine learning model utilizing time series data from major companies, including Amazon, Apple, Google, Nvidia, Meta, Tesla, Salesforce, Intel, and Microsoft. We explore the use of Long Short-Term Memory (LSTM) neural networks, to capture the temporal dynamics of stock prices. These models are designed to process sequential data, maintaining short term and long …