Computer Sciences Commons^™

Optimizing Uncertainty Quantification Of Vision Transformers In Deep Learning On Novel Ai Architectures, Erik Pautsch, John Li, Silvio Rizzi, George K. Thiruvathukal, Maria Pantoja

Computer Science: Faculty Publications and Other Works

Deep Learning (DL) methods have shown substantial efficacy in computer vision (CV) and natural language processing (NLP). Despite their proficiency, the inconsistency in input data distributions can compromise prediction reliability. This study mitigates this issue by introducing uncertainty evaluations in DL models, thereby enhancing dependability through a distribution of predictions. Our focus lies on the Vision Transformer (ViT), a DL model that harmonizes both local and global behavior. We conduct extensive experiments on the ImageNet-1K dataset, a vast resource with over a million images across 1,000 categories. ViTs, while competitive, are vulnerable to adversarial attacks, making uncertainty estimation crucial for …

Evaluating Methods For Improving Dnn Robustness Against Adversarial Attacks, Laureano Griffin

USF Tampa Graduate Theses and Dissertations

Deep learning has become more widespread as advances in the field continue. As aresult, making sure deep learning is safe has become a priority. A seemingly normal image with intentional pixel changes can cause a well-trained model to misclassify the image with high confidence. Those kinds of images are called adversarial attacks. Adversarial training has been developed to defend against adversarial attacks. This thesis evaluates different adversarial training methods against a variety of adversarial attacks. The key metrics for evaluation are classification accuracy and training time. This thesis also experiments with an improvement on an existing adversarial training method, the …

Semantic Lung Segmentation From Chest X-Ray Images Using Seg-Net Deep Cnn Model, Dathar Abas Hasan, Umed Hayder Jader

Polytechnic Journal

Implementing an accurate image segmentation to extract the lung shape from X-ray images is a vital step in designing a CAD system that diagnoses various types of chest diseases. Lung segmentation is a complex process due to the blurred regions that separate the lung area and the rest of the image. The conventional image segmentation techniques do not meet the ambitions to achieve precise lung segmentation. In this paper, we utilized the Seg-Net semantic segmentation model as a practical approach to distinguish the lung region pixels in X-ray images. The model involves an encoder network that extracts the data from …

Improving Semantic Document Classification Accuracy By Integrating Human-Crafted Knowledge, Zachary Weinfeld, Lubomir Stanchev

College of Engineering Summer Undergraduate Research Program

Document classification is a pivotal task in various domains, warranting the development of robust algorithms. Among these, the Bidirectional Encoder Representations from Transformers (BERT) algorithm, introduced by Google, has proven to perform well when fine-tuned for the task at hand. Leveraging transformer architecture, BERT demonstrates stellar language understanding capabilities. However, the integration of BERT with a range of techniques has shown potential for further enhancing classification accuracy. This work investigates several techniques that leverage semantic understanding to improve the performance of document classification models trained with BERT. Specifically, we explore three methods. First, we will balance corpuses afflicted by imbalanced …

Flacgec: A Chinese Grammatical Error Correction Dataset With Fine-Grained Linguistic Annotation, Hanyue Du, Yike Zhao, Qingyuan Tian, Jiani Wang, Lei Wang, Yunshi Lan, Xuesong Lu

Research Collection School Of Computing and Information Systems

Chinese Grammatical Error Correction (CGEC) has been attracting growing attention from researchers recently. In spite of the fact that multiple CGEC datasets have been developed to support the research, these datasets lack the ability to provide a deep linguistic topology of grammar errors, which is critical for interpreting and diagnosing CGEC approaches. To address this limitation, we introduce FlaCGEC, which is a new CGEC dataset featured with fine-grained linguistic annotation. Specifically, we collect raw corpus from the linguistic schema defined by Chinese language experts, conduct edits on sentences via rules, and refine generated samples manually, which results in 10k sentences …

Emotion-Aware Music Recommendation, Hieu Tran, Tuan Le, Anh Do, Tram Vu, Steven Bogaerts, Brian T. Howard

Computer Science Faculty publications

It is common to listen to songs that match one's mood. Thus, an AI music recommendation system that is aware of the user's emotions is likely to provide a superior user experience to one that is unaware. In this paper, we present an emotion-aware music recommendation system. Multiple models are discussed and evaluated for affect identification from a live image of the user. We propose two models: DRViT, which applies dynamic routing to vision transformers, and InvNet50, which uses involution. All considered models are trained and evaluated on the AffectNet dataset. Each model outputs the user's estimated valence and arousal …

Gpachov At Checkthat! 2023: A Diverse Multi-Approach Ensemble For Subjectivity Detection In News Articles, Georgi Pachov, Dimitar Dimitrov, Ivan Koychev, Preslav Nakov

Natural Language Processing Faculty Publications

The wide-spread use of social networks has given rise to subjective, misleading, and even false information on the Internet. Thus, subjectivity detection can play an important role in ensuring the objectiveness and the quality of a piece of information. This paper presents the solution built by the Gpachov team for the CLEF-2023 CheckThat! lab Task 2 on subjectivity detection. Three different research directions are explored. The first one is based on fine-tuning a sentence embeddings encoder model and dimensionality reduction. The second one explores a sample-efficient few-shot learning model. The third one evaluates fine-tuning a multilingual transformer on an altered …

Pattern-Of-Life Modeling With Automatic Dependent Surveillance-Broadcast (Ads-B), Sarah J. Bolton

Theses and Dissertations

This dissertation and research were sponsored by the Air Force Research Laboratory Layered Sensing Exploitation Branch (AFRL/RYA) to investigate the utility of using the data found within aircraft secondary radar to make predictions about aircraft characteristics and intent. The research focuses on making predictions on aircraft characteristics using only the kinetic data within one type of secondary radar, Automatic Dependent Surveillance-Broadcast (ADS-B), as a surrogate for primary radar. The results from this research provide a means to reduce the reliance on a type of aircraft tracking that is vulnerable to cyber attack and other integrity concerns.

Out-Of-Distribution Generalization Of Deep Learning To Illuminate Dark Protein Functional Space, Tian Cai

Dissertations, Theses, and Capstone Projects

Dark protein illumination is a fundamental challenge in drug discovery where majority human proteins are understudied, i.e. with only known protein sequence but no known small molecule binder. It's a major road block to enable drug discovery paradigm shift from single-targeted which looks to identify a single target and design drug to regulate the single target to multi-targeted in a Systems Pharmacology perspective. Diseases such as Alzheimer's and Opioid-Use-Disorder plaguing millions of patients call for effective multi-targeted approach involving dark proteins. Using limited protein data to predict dark protein property requires deep learning systems with OOD generalization capacity. Out-of-Distribution (OOD) …

Optimization And Application Of Graph Neural Networks, Shuo Zhang

Dissertations, Theses, and Capstone Projects

Graph Neural Networks (GNNs) are widely recognized for their potential in learning from graph-structured data and solving complex problems. However, optimal performance and applicability of GNNs have been an open-ended challenge. This dissertation presents a series of substantial advances addressing this problem. First, we investigate attention-based GNNs, revealing a critical shortcoming: their ignorance of cardinality information that impacts their discriminative power. To rectify this, we propose Cardinality Preserved Attention (CPA) models that can be applied to any attention-based GNNs, which exhibit a marked improvement in performance. Next, we introduce the Directional Node Pair (DNP) descriptor and the Robust Molecular Graph …

Learning Representations For Effective And Explainable Software Bug Detection And Fixing, Yi Li

Dissertations

Software has an integral role in modern life; hence software bugs, which undermine software quality and reliability, have substantial societal and economic implications. The advent of machine learning and deep learning in software engineering has led to major advances in bug detection and fixing approaches, yet they fall short of desired precision and recall. This shortfall arises from the absence of a 'bridge,' known as learning code representations, that can transform information from source code into a suitable representation for effective processing via machine and deep learning.

This dissertation builds such a bridge. Specifically, it presents solutions for effectively learning …

Sctiger: A Deep-Learning Method For Inferring Gene Regulatory Networks From Case Versus Control Scrna-Seq Datasets., Madison Dautle, Shaoqiang Zhang, Yong Chen

College of Science & Mathematics Departmental Research

Inferring gene regulatory networks (GRNs) from single-cell RNA-seq (scRNA-seq) data is an important computational question to find regulatory mechanisms involved in fundamental cellular processes. Although many computational methods have been designed to predict GRNs from scRNA-seq data, they usually have high false positive rates and none infer GRNs by directly using the paired datasets of case-versus-control experiments. Here we present a novel deep-learning-based method, named scTIGER, for GRN detection by using the co-differential relationships of gene expression profiles in paired scRNA-seq datasets. scTIGER employs cell-type-based pseudotiming, an attention-based convolutional neural network method and permutation-based significance testing for inferring GRNs among …

Deep Learning-Based Cad System For Predicting The Covid-19 X-Ray Images, Aqeel R. Talib, Hana’ M. Ali

Karbala International Journal of Modern Science

According to World Health Organization data, Coronavirus (COVID-19) has infected about 660, 378, 145 patients around the world. It is nonetheless difficult for physicians to detect COVID-19 infections out of CT or X-ray radiographs. Thus, several computer-aided diagnosis (CAD) systems based on deep learning and radiographs were developed to detect COVID-19 infections. However, the majority of approaches considered small datasets, which is ineligible to provide diverse COVID-19 radiographs. This work utilizes a massive number of X-ray radiographs, and compared standard CNN, DenseNet-121, and GoogLeNet for isolating COVID-19 infections out from normal and other pneumonia radiographs. The dataset in this work …

Generalizable Deep-Learning-Based Wireless Indoor Localization, Ali Owfi

All Theses

The growing interest in indoor localization has been driven by its wide range of applications in areas such as smart homes, industrial automation, and healthcare. With the increasing reliance on wireless devices for location-based services, accurate estimation of device positions within indoor environments has become crucial. Deep learning approaches have shown promise in leveraging wireless parameters like Channel State Information (CSI) and Received Signal Strength Indicator (RSSI) to achieve precise localization. However, despite their success in achieving high accuracy, these deep learning models suffer from limited generalizability, making them unsuitable for deployment in new or dynamic environments without retraining. To …

Visual Complexity Of The Time-Frequency Image Pinpoints The Epileptogenic Zone: An Unsupervised Deep-Learning Tool To Analyze Interictal Intracranial Eeg, Sarvagya Gupta

Graduate Masters Theses

Epilepsy, a prevalent neurological disorder characterized by recurrent seizures, continues to pose significant challenges in diagnosis and treatment, particularly among children. Despite substantial advancements in medical technology and treatment modalities, localization of the part of brain that causes seizures (Epileptogenic Zone) remains a difficult task. Intracranial EEG (iEEG) is often used to estimate the epileptogenic zone (EZ) in children with drugresistant epilepsy (DRE) and target it during surgery. Conventionally, iEEG signals are inspected in the time domain by human experts aiming to locate epileptiform activity.

Visual scrutiny of the iEEG time-frequency (TF) images can be an alternative way to review …

Invading The Integrity Of Deep Learning (Dl) Models Using Lsb Perturbation & Pixel Manipulation, Ashraful Tauhid

Theses and Dissertations

The use of deep learning (DL) models for solving classification and recognition-related problems are expanding at an exponential rate. However, these models are computationally expensive both in terms of time and resources. This imposes an entry barrier for low-profile businesses and scientific research projects with limited resources. Therefore, many organizations prefer to use fully outsourced trained models, cloud computing services, pre-trained models are available for download and transfer learning. This ubiquitous adoption of DL has unlocked numerous opportunities but has also brought forth potential threats to its prospects. Among the security threats, backdoor attacks and adversarial attacks have emerged as …

Generalizing Deep Learning Methods For Particle Tracing Using Transfer Learning, Shubham Gupta

All Graduate Theses and Dissertations, Spring 1920 to Summer 2023

Particle tracing is a very important method for scientific visualization of vector fields, but it is computationally expensive. Deep learning can be used to speed up particle tracing, but existing deep learning models are domain-specific. In this work, we present a methodology to generalize the use of deep learning for particle tracing using transfer learning. We demonstrate the performance of our approach through a series of experimental studies that address the most common simulation design scenarios: varying time span, Reynolds number, and problem geometry. The results show that our methodology can be effectively used to generalize and accelerate the training …

Geospatial Wildfire Risk Prediction Using Deep Learning, Abner Alberto Benavides

Electronic Theses, Projects, and Dissertations

This report introduces a thorough analysis of wildfire prediction using satellite imagery by applying deep learning techniques. To find wildfire-prone geographical data, we use U-Net, a convolutional neural network known for its effectiveness in biomedical image segmentation. The input to the model is the Sentinel-2 multispectral images to supply a complete view of the terrain features.

We evaluated the wildfire risk prediction model’s performance using several metrics. The model showed high accuracy, with a weighted average F1 score of 0.91 and an AUC-ROC score of 0.972. These results suggest that the model is exceptionally good at predicting the location of …

Spatial And Temporal Agnostic Deep-Learning Based Radio Fingerprinting, Fahmida Afrin

School of Computing: Dissertations, Theses, and Student Research

Radio fingerprinting is a technique that validates wireless devices based on their unique radio frequency (RF) signals. This method is highly feasible because RF signals carry distinct hardware variations introduced during manufacturing. The security and trustworthiness of current and future wireless networks heavily rely on radio fingerprinting. In addition to identifying individual devices, it can also differentiate mission-critical targets. Despite significant efforts in the literature, existing radio fingerprinting methods require improved robustness, scalability, and resilience. This study focuses on the challenges of spatial-temporal variations in the wireless environment. Many prior approaches overlook the complex numerical structure of the in-phase and …

Using Deep Learning For Encrypted Traffic Analysis Of Amazon Echo, Surendra Pathak

Theses and Dissertations

The adoption of the Amazon Echo family of devices in modern homes has become very widespread at the current time, with hundreds of millions of devices sold. Moreover, the global smart speaker market size is growing vigorously and is projected to continue to bigger. Smart speakers allow users hands-free interaction by allowing voice control, promoting human-computer interaction to greater avenues. Though smart speaker can be useful assistant, it has some serious security concerns that need to be studied. In this study, an analysis of the security and privacy concerns of smart speakers is presented along with a passive attack, namely …

Extending The Convolution In Graph Neural Networks To Solve Materials Science And Node Classification Problems, Steph-Yves Mike Louis

Theses and Dissertations

The usage of graph to represent one's data in machine learning has grown in popularity in both academia and the industry due to its inherent benefits. With its flexible nature and immediate translation to real life observed objects, graph representation had a considerable contribution in advancing the state-of-the-art performance of machine learning in materials.

In this dissertation proposal, we discuss how machines can learn from graph encoded data and provide excellent results through graph neural networks (GNN). Notably, we focus our adaptation of graph neural networks on three tasks: predicting crystal materials properties, nullifying the negative impact of inferior graph …

Predicting Material Structures And Properties Using Deep Learning And Machine Learning Algorithms, Yuqi Song

Theses and Dissertations

Discovering new materials and understanding their crystal structures and chemical properties are critical tasks in the material sciences. Although computational methodologies such as Density Functional Theory (DFT), provide a convenient means for calculating certain properties of materials or predicting crystal structures when combined with search algorithms, DFT is computationally too demanding for structure prediction and property calculation for most material families, especially for those materials with a large number of atoms. This dissertation aims to address this limitation by developing novel deep learning and machine learning algorithms for effective prediction of material crystal structures and properties. Our data-driven machine learning …

Duplicate Bug Report Detection: How Far Are We?, Ting Zhang, Donggyun Han, Venkatesh Vinayakarao, Ivana Clairine Irsan, Bowen Xu, Thung Ferdian, David Lo, Lingxiao Jiang

Research Collection School Of Computing and Information Systems

Many Duplicate Bug Report Detection (DBRD) techniques have been proposed in the research literature. The industry uses some other techniques. Unfortunately, there is insufficient comparison among them, and it is unclear how far we have been. This work fills this gap by comparing the aforementioned techniques. To compare them, we first need a benchmark that can estimate how a tool would perform if applied in a realistic setting today. Thus, we first investigated potential biases that affect the fair comparison of the accuracy of DBRD techniques. Our experiments suggest that data age and issue tracking system choice cause a significant …

A Novel Approach To Extending Music Using Latent Diffusion, Keon Roohparvar, Franz J. Kurfess

Master's Theses

Using deep learning to synthetically generate music is a research domain that has gained more attention from the public in the past few years. A subproblem of music generation is music extension, or the task of taking existing music and extending it. This work proposes the Continuer Pipeline, a novel technique that uses deep learning to take music and extend it in 5 second increments. It does this by treating the musical generation process as an image generation problem; we utilize latent diffusion models (LDMs) to generate spectrograms, which are image representations of music. The Continuer Pipeline is able to …

Deep Learning Recommendations For The Acl2 Interactive Theorem Prover, Robert K. Thompson, Robert K. Thompson

Master's Theses

Due to the difficulty of obtaining formal proofs, there is increasing interest in partially or completely automating proof search in interactive theorem provers. Despite being a theorem prover with an active community and plentiful corpus of 170,000+ theorems, no deep learning system currently exists to help automate theorem proving in ACL2. We have developed a machine learning system that generates recommendations to automatically complete proofs. We show that our system benefits from the copy mechanism introduced in the context of program repair. We make our system directly accessible from within ACL2 and use this interface to evaluate our system in …

Detecting Ai Generated Text Using Neural Networks, Jesus Guerrero

Masters Theses (Archived)

For humans, distinguishing machine generated text from human written text is men- tally taxing and slow. NLP models have been created to do this more effectively and faster. But, what if some adversarial changes have been added to the machine generated text? This thesis discusses this issue and text detectors in general.
/="/">
/="/">The primary goal of this thesis is to describe the current state of text detectors in research and to discuss a key adversarial issue in modern NLP transformers. To describe the current state of text detectors a Systematic Literature Review was done on 50 relevant papers …

Boosting Adversarial Training Using Robust Selective Data Augmentation, Bader Rasheed, Asad Masood Khattak, Adil Khan, Stanislav Protasov, Muhammad Ahmad

All Works

Artificial neural networks are currently applied in a wide variety of fields, and they are near to achieving performance similar to humans in many tasks. Nevertheless, they are vulnerable to adversarial attacks in the form of a small intentionally designed perturbation, which could lead to misclassifications, making these models unusable, especially in applications where security is critical. The best defense against these attacks, so far, is adversarial training (AT), which improves the model’s robustness by augmenting the training data with adversarial examples. In this work, we show that the performance of AT can be further improved by employing the neighborhood …

Enhancing Health Tweet Classification: An Evaluation Of Transformer-Based Models For Comprehensive Analysis, Foram Pankajbhai Patel

Computer Science and Engineering Theses - Archive

The task of health tweet classification entails identifying whether a given tweet is health-related or not. While existing research in this area has made significant progress in classifying tweets into specific sub-domains of health, such as mental health, COVID-19, or specific diseases, there is a need for a more comprehensive approach that considers a broader range of health-related topics. This thesis addresses this need by proposing a diverse and comprehensive dataset that includes various existing health-related datasets, data collected through a keyword-based approach, and manually annotated data. However, the use of health-related keywords in a figurative or non-health context poses …

Automated Classification Of Pectinodon Bakkeri Teeth Images Using Machine Learning, Jacob A. Bahn

MS in Computer Science Project Reports

Microfossil dinosaur teeth are studied by paleontologists in order to better under- stand dinosaurs. Currently, tooth classification is a long, manual, error-ridden process. Deep learning offers a solution that allows for an automated way of classifying images of these microfossil teeth. In this thesis, we aimed to use deep learning in order to develop an automated approach for classifying images of Pectinodon bakkeri teeth. The proposed model was trained using a custom topology and it classified the images based on clusters created via K-Means. The model had an accuracy of 71%, a precision of 71%, a recall of 70.5%, and …

Chatgpt As Metamorphosis Designer For The Future Of Artificial Intelligence (Ai): A Conceptual Investigation, Amarjit Kumar Singh (Library Assistant), Dr. Pankaj Mathur (Deputy Librarian)

Library Philosophy and Practice (e-journal)

Abstract

Purpose: The purpose of this research paper is to explore ChatGPT’s potential as an innovative designer tool for the future development of artificial intelligence. Specifically, this conceptual investigation aims to analyze ChatGPT’s capabilities as a tool for designing and developing near about human intelligent systems for futuristic used and developed in the field of Artificial Intelligence (AI). Also with the helps of this paper, researchers are analyzed the strengths and weaknesses of ChatGPT as a tool, and identify possible areas for improvement in its development and implementation. This investigation focused on the various features and functions of ChatGPT that …