Physical Sciences and Mathematics Commons^™

Transformer-Based Deep Learning Prediction Of 10-Degree Humphrey Visual Field Tests From 24-Degree Data, Min Shi, Anagha Lokhande, Yu Tian, Yan Luo, Mohammad Eslami, Saber Kazeminasab, Tobias Elze, Lucy Shen, Louis Pasquale, Sarah Wellik, Carlos Gustavo De Moraes, Jonathan Myers, Nazlee Zebardast, David Friedman, Michael Boland, Mengyu Wang

Wills Eye Hospital Papers

PURPOSE: To predict 10-2 Humphrey visual fields (VFs) from 24-2 VFs and associated non-total deviation features using deep learning.

METHODS: We included 5189 reliable 24-2 and 10-2 VF pairs from 2236 patients, and 28,409 reliable pairs of macular OCT scans and 24-2 VF from 19,527 eyes of 11,560 patients. We developed a transformer-based deep learning model using 52 total deviation values and nine VF test features to predict 68 10-2 total deviation values. The mean absolute error, root mean square error, and the R2 were evaluation metrics. We further evaluated whether the predicted 10-2 VFs can improve the structure-function relationship …

High Prevalence Of Artifacts In Optical Coherence Tomography With Adequate Signal Strength, Wei-Chun Lin, Aaron Coyner, Charles Amankwa, Abigail Lucero, Gadi Wollstein, Joel Schuman, Hiroshi Ishikawa

Wills Eye Hospital Papers

PURPOSE: This study aims to investigate the prevalence of artifacts in optical coherence tomography (OCT) images with acceptable signal strength and evaluate the performance of supervised deep learning models in improving OCT image quality assessment.

METHODS: We conducted a retrospective study on 4555 OCT images from 546 patients, with each image having an acceptable signal strength (≥6). A comprehensive analysis of prevalent OCT artifacts was performed, and five pretrained convolutional neural network models were trained and tested to infer images based on quality.

RESULTS: Our results showed a high prevalence of artifacts in OCT images with acceptable signal strength. Approximately …

Deep Representation Learning For Time Series Forecasting, Gerald Woo

Dissertations and Theses Collection (Open Access)

Time series forecasting has critical applications across business and scien- tific domains, such as demand forecasting, capacity planning and management, and anomaly detection. Being able to predict the future yields immense value, allowing us to make downstream decisions with more confidence. Deep learning for time series forecasting is a burgeoning area of research, moving away from simple linear models found in classical time series analysis literature, towards more expressive, data hungry neural network architectures.

In this thesis, we develop methods leveraging deep representation learning for time series forecasting, from exploring neural network architecture designs which encode inductive biases specific to …

Interpretation Models For Prostate Lesion: Detecting, Explaining, And Understanding., Mehmet Akif Gulum

Electronic Theses and Dissertations

Prostate cancer is a major public health concern, affecting millions of men worldwide. While early detection and treatment of prostate cancer is critical for improving patient outcomes, the detection of prostate lesions is even more important for timely intervention and management of the disease. Prostate lesions are abnormal growths or lumps within the prostate gland, which may or may not be cancerous. The timely detection and accurate diagnosis of prostate lesions is crucial for effective treatment and management of the disease. In recent years, deep learning models have shown promise in accurately detecting and characterizing prostate lesions using advanced imaging …

Checklist For Reproducibility Of Deep Learning In Medical Imaging, Mana Moassefi, Yashbir Singh, Gian Marco Conte, Bardia Khosravi, Pouria Rouzrokh, Sanaz Vahdati, Nabile Safdar, Linda Moy, Felipe Kitamura, Amilcare Gentili, Paras Lakhani, Nina Kottler, Safwan Halabi, Joseph Yacoub, Yuankai Hou, Khaled Younis, Bradley Erickson, Elizabeth Krupinski, Shahriar Faghani

Department of Radiology Faculty Papers

The application of deep learning (DL) in medicine introduces transformative tools with the potential to enhance prognosis, diagnosis, and treatment planning. However, ensuring transparent documentation is essential for researchers to enhance reproducibility and refine techniques. Our study addresses the unique challenges presented by DL in medical imaging by developing a comprehensive checklist using the Delphi method to enhance reproducibility and reliability in this dynamic field. We compiled a preliminary checklist based on a comprehensive review of existing checklists and relevant literature. A panel of 11 experts in medical imaging and DL assessed these items using Likert scales, with two survey …

A Comprehensive Dataset For Arabic Word Sense Disambiguation, Sanaa Kaddoura, Reem Nassar

All Works

This data paper introduces a comprehensive dataset tailored for word sense disambiguation tasks, explicitly focusing on a hundred polysemous words frequently employed in Modern Standard Arabic. The dataset encompasses a diverse set of senses for each word, ranging from 3 to 8, resulting in 367 unique senses. Each word sense is accompanied by contextual sentences comprising ten sentence examples that feature the polysemous word in various contexts. The data collection resulted in a dataset of 3670 samples. Significantly, the dataset is in Arabic, which is known for its rich morphology, complex syntax, and extensive polysemy. The data was meticulously collected …

Multi-Label Voice Disorder Classification Using Raw Waveforms, Gökay Di̇şken

Turkish Journal of Electrical Engineering and Computer Sciences

Automated voice disorder systems that distinguish pathological voices from healthy ones have been developed with the aid of machine learning methods. Both clinicians and patients can benefit from these systems as they provide many advantages, compared to the invasive techniques. These systems can produce binary (healthy/pathological) or multi-class (healthy/selected pathologies) decisions. However, multiple disorders might exist in an individual’s voice. Multi-label classification should be considered in such cases. By this time, only a single report is available on this topic, where hand-crafted features were used, and a data augmentation technique was utilized to overcome class imbalances. In this study, a …

Enrichment Of Turkish Question Answering Systems Using Knowledge Graphs, Okan Çi̇ftçi̇, Fati̇h Soygazi̇, Selma Teki̇r

Turkish Journal of Electrical Engineering and Computer Sciences

Recent capabilities of large language models (LLMs) have transformed many tasks in Natural Language Processing (NLP), including question answering. The state-of-the-art systems do an excellent job of responding in a relevant, persuasive way but cannot guarantee factuality. Knowledge graphs, representing facts as triplets, can be valuable for avoiding errors and inconsistencies with real-world facts. This work introduces a knowledge graph-based approach to Turkish question answering. The proposed approach aims to develop a methodology capable of drawing inferences from a knowledge graph to answer complex multihop questions. We construct the Beyazperde Movie Knowledge Graph (BPMovieKG) and the Turkish Movie Question Answering …

Network Intrusion Detection Based On Machine Learning Strategies: Performance Comparisons On Imbalanced Wired, Wireless, And Software-Defined Networking (Sdn) Network Traffics, Hi̇lal Hacilar, Zafer Aydin, Vehbi̇ Çağri Güngör

Turkish Journal of Electrical Engineering and Computer Sciences

The rapid growth of computer networks emphasizes the urgency of addressing security issues. Organizations rely on network intrusion detection systems (NIDSs) to protect sensitive data from unauthorized access and theft. These systems analyze network traffic to detect suspicious activities, such as attempted breaches or cyberattacks. However, existing studies lack a thorough assessment of class imbalances and classification performance for different types of network intrusions: wired, wireless, and software-defined networking (SDN). This research aims to fill this gap by examining these networks’ imbalances, feature selection, and binary classification to enhance intrusion detection system efficiency. Various techniques such as SMOTE, ROS, ADASYN, …

A New Optimization Approach Based On Neural Architecture Search To Enhance Deep U-Net For Efficient Road Segmentation, Narges Saeedizadeh, Seyed Mohammad Jafar Jalali, Burhan Khan, Parham Mohsenzadeh Kebria, Shady Mohamed

Research outputs 2022 to 2026

Neural Architecture Search (NAS) has significantly improved the accuracy of image classification and segmentation. However, these methods concentrate on finding segmentation structures for natural or medical applications. In this study, we introduce a NAS approach based on gradient optimization to identify ideal cell designs for road segmentation. To the best of our knowledge, this work represents the first application of gradient-based NAS to road extraction. Taking insight from the U-Net model and its successful variations in different image segmentation tasks, we propose NAS-enhanced U-Net, illustrated by an equal number of cells in both encoder and decoder levels. While cross-entropy combined …

Label-Free Surface-Enhanced Raman Spectroscopy Coupled With Machine Learning Algorithms In Pathogenic Microbial Identification: Current Trends, Challenges, And Perspectives, Jia Wei Tang, Quan Yuan, Xin Ru Wen, Muhammad Usman, Alfred Chin Yen Tay, Liang Wang

Research outputs 2022 to 2026

Infectious diseases caused by microbial pathogens remain a primary contributor to global health burdens. Prompt control and effective prevention of these pathogens are critical for public health and medical diagnostics. Conventional microbial detection methods suffer from high complexity, low sensitivity, and poor selectivity. Therefore, developing rapid and reliable methods for microbial pathogen detection has become imperative. Surface-enhanced Raman Spectroscopy (SERS), as an innovative non-invasive diagnostic technique, holds significant promise in pathogenic microorganism detection due to its rapid, reliable, and cost-effective advantages. This review comprehensively outlines the fundamental theories of Raman Spectroscopy (RS) with a focus on label-free SERS strategy, reporting …

On The Sustainability Of Deep Learning Projects: Maintainers' Perspective, Junxiao Han, Jiakun Liu, David Lo, Chen Zhi, Yishan Chen, Shuiguang Deng

Research Collection School Of Computing and Information Systems

Deep learning (DL) techniques have grown in leaps and bounds in both academia and industry over the past few years. Despite the growth of DL projects, there has been little study on how DL projects evolve, whether maintainers in this domain encounter a dramatic increase in workload and whether or not existing maintainers can guarantee the sustained development of projects. To address this gap, we perform an empirical study to investigate the sustainability of DL projects, understand maintainers' workloads and workloads growth in DL projects, and compare them with traditional open-source software (OSS) projects. In this regard, we first investigate …

Deep Learning In Reproducing Kernel Banach Spaces, Mingsong Yan

Mathematics & Statistics Theses & Dissertations

Deep learning has achieved immense success in the past decade. The goal of this dissertation is to understand deep learning through the framework of reproducing kernel Banach spaces (RKBSs), which were originally proposed for promoting sparse solutions. We begin by considering learning problems in a general functional setting, and establishing explicit and data-dependent representer theorems for both minimal norm interpolation (MNI) problems and regularization problems. These theorems provide a crucial foundation for the subsequent results derived for both sparse learning and deep learning. Next, we investigate the essential properties of RKBSs capable of encouraging sparsity in learning solutions. With the …

Harnessing Social Media For Disaster Response: Intelligent Identification Of Reliable Rescue Requests During Hurricanes, Wael Khallouli

Engineering Management & Systems Engineering Theses & Dissertations

Hurricanes pose a significant threat to both human lives and infrastructure. Decision-makers face substantial challenges during such events, as they must act quickly to address victims’ needs. Social media platforms provide a valuable source for quick and real-time information. Recent hurricane events have shown that people turn to social media to call for help when official communication channels, such as 911, are overwhelmed. However, extracting actionable information from the massive number of messages posted on social media is challenging. Furthermore, verifying social media messages posted by the public is a critical concern for disaster response practitioners, making them hesitant to …

Comparative Analysis Of Hate Speech Detection: Traditional Vs. Deep Learning Approaches, Haibo Pen, Nicole Anne Huiying Teo, Zhaoxia Wang

Research Collection School Of Computing and Information Systems

Detecting hate speech on social media poses a significant challenge, especially in distinguishing it from offensive language, as learning-based models often struggle due to nuanced differences between them, which leads to frequent misclassifications of hate speech instances, with most research focusing on refining hate speech detection methods. Thus, this paper seeks to know if traditional learning-based methods should still be used, considering the perceived advantages of deep learning in this domain. This is done by investigating advancements in hate speech detection. It involves the utilization of deep learning-based models for detailed hate speech detection tasks and compares the results with …

Transformer Models With Explainability For It Telemetry And Business Events, Shiau Hong Lim, Laura Wynter

Research Collection School Of Computing and Information Systems

Temporal event data are commonly encountered in software applications across a wide range of domains, from IT telemetry and system logs to business process automation. Temporal event data in software applications carry information in various forms: both structured and unstructured, and with both regular and irregular occurrence frequency, making the representation of temporal event data challenging. Further chal-lenges come from the diversity in terms of the scale and volume of events that need to be summarized by the representation. We propose a general and unified approach to handle temporal event data for the purpose of learning a predictive transformer model. …

Unified Training Of Universal Time Series Forecasting Transformers, Gerald Woo, Chenghao Liu, Akshat Kumar, Caiming Xiong, Silvio Savarese, Doyen Sahoo

Research Collection School Of Computing and Information Systems

Deep learning for time series forecasting has traditionally operated within a one-model-per-dataset framework, limiting its potential to leverage the game-changing impact of large pre-trained models. The concept of universal forecasting, emerging from pre-training on a vast collection of time series datasets, envisions a single Large Time Series Model capable of addressing diverse downstream forecasting tasks. However, constructing such a model poses unique challenges specific to time series data: i) cross-frequency learning, ii) accommodating an arbitrary number of variates for multivariate time series, and iii) addressing the varying distributional properties inherent in large-scale data. To address these challenges, we present novel …

Hardening Interpretable Deep Learning Systems: Investigating Adversarial Threats And Defenses, Eldor Abdukhamidov, Mohammed Abuhamad, Simon S. Woo, Eric Chan-Tin, Tamer Abuhmed

Computer Science: Faculty Publications and Other Works

Deep learning methods have gained increasing attention in various applications due to their outstanding performance. For exploring how this high performance relates to the proper use of data artifacts and the accurate problem formulation of a given task, interpretation models have become a crucial component in developing deep learning-based systems. Interpretation models enable the understanding of the inner workings of deep learning models and offer a sense of security in detecting the misuse of artifacts in the input data. Similar to prediction models, interpretation models are also susceptible to adversarial inputs. This work introduces two attacks, AdvEdge and AdvEdge + …

Efficient Machine Learning On Scientific Data Using Bayesian Optimization, Rui Xin

Theses and Dissertations

Deep Learning is pivotal in advancing data analysis across various scientific fields, from genomics to materials discovery. Despite its widespread use, efficiently learning from limited data and operating under resource constraints remains a significant challenge, often limiting its full potential in environments where data is scarce or resources are restricted. This dissertation explores Active Learning and Automated Machine Learning (AutoML) powered by Bayesian Optimization to enhance the efficiency of machine learning across multiple disciplines. It focuses on algorithm optimization and data management through three interconnected studies. In the first study, we investigate how data management technique - active learning helps …

Face Mask Detection Based On Deep Learning: A Review, Shahad Fadhil Abbas, Shaimaa Hameed Shaker, Firas. A. Abdullatif

Journal of Soft Computing and Computer Applications

The coronavirus disease 2019 outbreak caused widespread disruption. The World Health Organization has recommended wearing face masks, along with other public health measures, such as social distancing, following medical guidelines, and thermal scanning, to reduce transmission, reduce the burden on healthcare systems, and protect population groups. However, wearing a mask, which acts as a barrier or shield to reduce transmission of infection from infected individuals, hides most facial features, such as the nose, mouth, and chin, on which face detection systems depend, which leads to the weakness of these systems. This paper aims to provide essential insights for researchers and …

Strangeness Detection From Crowded Video Scenes By Hand-Crafted And Deep Learning Features, Ali A. Hussan, Shaimaa H. Shaker, Akbas Ezaldeen Ali

Journal of Soft Computing and Computer Applications

Video anomaly detection is one of the trickiest issues in intelligent video surveillance because of the complexity of real data and the hazy definition of anomalies. Since abnormal occurrences typically seem different from normal events and move differently. The global optical flow was determined with the maximum accuracy and speed using the Farneback approach for calculating the magnitudes. Two approaches have been used in this study to detect strangeness in the video. These approaches are Deep Learning (DL) and manuality. The first method uses the activity map's development of entropy to detect the oddity in the video using a particular …

A Comprehensive Analysis Of Deep Learning And Swarm Intelligence Techniques To Enhance Vehicular Ad-Hoc Network Performance, Hussein K. Abdul Atheem, Israa T. Ali, Faiz A. Al Alawy

Journal of Soft Computing and Computer Applications

The primary elements of Intelligent Transportation Systems (ITSs) have become Vehicular Ad-hoc NETworks (VANETs), allowing communication between the infrastructure environment and vehicles. The large amount of data gathered by connected vehicles has simplified how Deep Learning (DL) techniques are applied in VANETs. DL is a subfield of artificial intelligence that provides improved learning algorithms able to analyzing and process complex and heterogeneous data. This study explains the power of DL in VANETs, considering applications like decision-making, vehicle localization, anomaly detection, traffic prediction and intelligent routing, various types of DL, including Recurrent Neural Networks (RNNs), and Convolutional Neural Networks (CNNs) are …

Procedural Pre-Training For Visual Recognition, Connor S. Anderson

Theses and Dissertations

Deep learning models can perform many tasks very capably, provided they are trained correctly. Usually, this requires a large amount of data. Pre-training refers to a process of creating a strong initial model by first training it on a large-scale dataset. Such a model can then be adapted to many different tasks, while only requiring a comparatively small amount of task-specific training data. Pre-training is the standard approach in most computer vision scenarios, but it's not without drawbacks. Aside from the cost and effort involved in collecting large pre-training datasets, such data may also contain unwanted biases, violations of privacy, …

Sre-Yolov8: : An Improved Uav Object Detection Model Utilizing Swin Transformer And Re-Fpn, Jun Li, Jiajie Zhang, Yanhua Shao, Feng Liu

Faculty, Staff and Student Publications

To tackle the intricate challenges associated with the low detection accuracy of images taken by unmanned aerial vehicles (UAVs), arising from the diverse sizes and types of objects coupled with limited feature information, we present the SRE-YOLOv8 as an advanced method. Our method enhances the YOLOv8 object detection algorithm by leveraging the Swin Transformer and a lightweight residual feature pyramid network (RE-FPN) structure. Firstly, we introduce an optimized Swin Transformer module into the backbone network to preserve ample global contextual information during feature extraction and to extract a broader spectrum of features using self-attention mechanisms. Subsequently, we integrate a Residual …

Case Studies For Energy Efficient Machine Learning Inference Acceleration, Recep Erol

Theses and Dissertations

The advancements in machine learning, deep learning and AI have yielded remarkable tools and innovations, but certain groups face barriers preventing their utilization of these technologies. This research identifies and categorizes these barriers, focusing on three distinct groups: those lacking computational power, seeking to deploy models across multiple devices, and struggling with optimization challenges in high-performance computing centers. The study highlights the disconnect between academia's proposed solutions and their practical integration within industries and research centers, emphasizing the lack of convenience and integration among existing tools. To bridge this gap, this research offers a multifaceted approach. Firstly, it introduces publicly …

Achieving Domain-Independent Certified Robustness Via Knowledge Continuity, Alan Wenyuan Sun

Computer Science Senior Theses

We present knowledge continuity, a novel definition inspired by Lipschitz continuity which aims to certify the robustness of neural networks across input domains (such as continuous and discrete domains in vision and language, respectively). Most existing approaches that seek to certify robustness, especially Lipschitz continuity, lie within the continuous domain with norm and distribution-dependent guarantees. In contrast, our proposed definition yields certification guarantees that depend only on the loss function and the intermediate learned metric spaces of the neural network. These bounds are independent of domain modality, norms, and distribution. We further demonstrate that the expressiveness of a model …

Emotional Regulation On Modulating Associations Between Depression And Physical Activity As Characterized Via Deep Learning, Franklin Ye Ruan

Computer Science Senior Theses

Emotional regulation and physical activity are known to be associated with depression; however, a deeper understanding of how emotional regulation may strengthen or weaken the bonds between depressive symptoms and physical activity may aid clinicians and researchers in developing cognitive behavioral therapy (CBT) for those adversely affected by depression. As part of the Tracking Depression Study, this analysis uses data collected from 306 participants diagnosed with Major Depressive Disorder. To study their behavior, we analyze actigraphy data, or longitudinal physical activity intensity data, as it relates to depression severity, quantified by the daily PHQ-9 questionnaires. We study these associations through …

Bagging Improves The Performance Of Deep Learning-Based Semantic Segmentation With Limited Labeled Images: A Case Study Of Crop Segmentation For High-Throughput Plant Phenotyping, Yinglun Zhan, Yuzhen Zhou, Geng Bai, Yufeng Ge

Department of Statistics: Faculty Publications

Advancements in imaging, computer vision, and automation have revolutionized various fields, including field-based high-throughput plant phenotyping (FHTPP). This integration allows for the rapid and accurate measurement of plant traits. Deep Convolutional Neural Networks (DCNNs) have emerged as a powerful tool in FHTPP, particularly in crop segmentation—identifying crops from the background—crucial for trait analysis. However, the effectiveness of DCNNs often hinges on the availability of large, labeled datasets, which poses a challenge due to the high cost of labeling. In this study, a deep learning with bagging approach is introduced to enhance crop segmentation using high-resolution RGB images, tested on the …

Text-To-Sql: A Methodical Review Of Challenges And Models, Ali Buğra Kanburoğlu, Faik Boray Tek

Turkish Journal of Electrical Engineering and Computer Sciences

This survey focuses on Text-to-SQL, automated translation of natural language queries into SQL queries. Initially, we describe the problem and its main challenges. Then, by following the PRISMA systematic review methodology, we survey the existing Text-to-SQL review papers in the literature. We apply the same method to extract proposed Text-to-SQL models and classify them with respect to used evaluation metrics and benchmarks. We highlight the accuracies achieved by various models on Text-to-SQL datasets and discuss execution-guided evaluation strategies. We present insights into model training times and implementations of different models. We also explore the availability of Text-to-SQL datasets in non-English …

Dpafy-Gcaps: Denoising Patch-And-Amplify Gabor Capsule Network For The Recognition Of Gastrointestinal Diseases, Henrietta Adjei Pokuaa, Adeboya Felix Adekoya, Benjamin Asubam Weyori, Owusu Nyarko-Boateng

Turkish Journal of Electrical Engineering and Computer Sciences

Deep learning (DL) models have performed tremendously well in image classification. This good performance can be attributed to the availability of massive data in most domains. However, some domains are known to have few datasets, especially the health sector. This makes it difficult to develop domain-specific high-performing DL algorithms for these fields. The field of health is critical and requires accurate detection of diseases. In the United States Gastrointestinal diseases are prevalent and affect 60 to 70 million people. Ulcerative colitis, polyps, and esophagitis are some gastrointestinal diseases. Colorectal polyps is the third most diagnosed malignancy in the world. This …