Digital Commons Network^™

Impact Of Spatial Diversity And Subject Variability On Wifi-Based Human Activity Recognition, Amany Elkelany, Robert J. Ross, Susan Mckeever

SAML-25 Workshop on Statistical and Machine Learning

In recent years, WiFi-based Human Activity Recognition (HAR) has gained substantial attention due to the ubiquity of WiFi infrastructure and advancements in wireless communication. Unlike camera-based systems that raise privacy concerns or wearable sensors that require user compliance, WiFi-based HAR provides a noninvasive and practical alternative that operates seamlessly with existing infrastructure. WiFi-based HAR leverages fluctuations in wireless signals, particularly Channel State Information (CSI), to passively detect and classify human activities. WiFi-based HAR models often achieve high accuracy in a single environment but suffer significant performance drops when applied to new environments due to variations in spatial settings, human movement, …

Deep Learning Based Rapid X-Ray Fluorescence Signal Extraction And Image Reconstruction For Preclinical Benchtop X-Ray Fluorescence Computed Tomography Applications, Amrit Kaphle, Sandun Jayarathna, Sang Hyun Cho

Faculty, Staff and Student Publications

Recent research advances have resulted in an experimental benchtop X-ray fluorescence computed tomography (XFCT) system that likely meets the imaging dose/scan time constraints for benchtop XFCT imaging of live mice injected with gold nanoparticles (GNPs). For routine in vivo benchtop XFCT imaging, however, additional challenges, most notably the need for rapid/near-real-time handling of X-ray fluorescence (XRF) signal extraction and XFCT image reconstruction, must be successfully addressed. Here we propose a novel end-to-end deep learning (DL) framework that integrates a one-dimensional convolutional neural network (1D CNN) for rapid XRF signal extraction with a U-Net model for XFCT image reconstruction. We trained …

Cross-Modality Learning For Predicting Ihc Biomarkers From H&E-Stained Whole-Slide Images, Amit Das

Computer Science Senior Theses

Hematoxylin and Eosin (H&E) staining is a cornerstone of pathological analysis, offering reliable visualization of cellular morphology and tissue architecture for cancer diagnosis, subtyping, and grading. Immunohistochemistry (IHC) staining, an important ancillary study, provides molecular insights by detecting specific proteins within tissues, enhancing diagnostic accuracy, and improving treatment planning. However, IHC staining is costly, time-consuming, and resource-intensive, requiring specialized expertise. To address these limitations, this study proposes HistoStainAlign, a novel deep learning framework that predicts IHC staining patterns directly from H&E whole-slide images (WSIs) by learning joint representations of morphological and molecular features. The framework integrates paired H&E and IHC …

Minifying Deep Denoising Networks With Knowledge Distillation, Antonio L. Rozzi

Master's Theses

Hearing loss is a prevalent condition, affecting hundreds of millions globally, with a higher incidence among older adults. While hearing aids are the standard treatment, the majority of those who could benefit from hearing aids choose not to wear them, attributing this decision in large part to their inability to perform well in conversations in large groups and in noisy situations. To date, no denoising systems on commercial hearing aids are able to improve speech intelligibility. Recent advances in artificial intelligence research have shown that large deep-learning models can in fact improve speech intelligibility by removing background noise from audio. …

Learning Spatio-Temporal Dynamics For Trajectory Recovery Via Time-Aware Transformer, Tian Sun, Yuqi Chen, Baihua Zheng, Weiwei Sun

Research Collection School Of Computing and Information Systems

In real-world applications, GPS trajectories often suffer from low sampling rates, with large and irregular intervals between consecutive GPS points. This sparse characteristic presents challenges for their direct use in GPS-based systems. This paper addresses the task of map-constrained trajectory recovery, aiming to enhance trajectory sampling rates of GPS trajectories. Previous studies commonly adopt a sequence-to-sequence framework, where an encoder captures the trajectory patterns and a decoder reconstructs the target trajectory. Within this framework, effectively representing the road network and extracting relevant trajectory features are crucial for overall performance. Despite advancements in these models, they fail to fully leverage the …

Linking Symptom Inventories Using Semantic Textual Similarity, Eamonn Kennedy, Shashank Vadlamani, Hannah M Lindsey, Kelly S Peterson, Kristen Dams O'Connor, Ronak Agarwal, Houshang H Amiri, Raeda K Andersen, Talin Babikian, David A Baron, Erin D Bigler, Karen Caeyenberghs, Lisa Delano-Wood, Seth G Disner, Ekaterina Dobryakova, Blessen C Eapen, Rachel M Edelstein, Carrie Esopenko, Helen M Genova, Elbert Geuze, Naomi J Goodrich-Hunsaker, Jordan Grafman, Asta K Håberg, Cooper B Hodges, Kristen R Hoskinson, Elizabeth S Hovenden, Andrei Irimia, Neda Jahanshad, Ruchira M Jha, Finian Keleher, Kimbra Kenney, Inga K Koerte, Spencer W Liebel, Abigail Livny, Marianne Løvstad, Sarah L Martindale, Jeffrey E Max, Andrew R Mayer, Timothy B Meier, Deleene S Menefee, Abdalla Z Mohamed, Stefania Mondello, Martin M Monti, Rajendra A Morey, Virginia Newcombe, Mary R Newsome, Alexander Olsen, Nicholas J Pastorek, Mary Jo Pugh, Adeel Razi, Jacob E Resch, Jared A Rowland, Kelly Russell, Nicholas P Ryan, Randall S Scheibel, Adam T Schmidt, Gershon Spitz, Jaclyn A Stephens, Assaf Tal, Leah D Talbert, Maria Carmela Tartaglia, Brian A Taylor, Sophia I Thomopoulos, Maya Troyanskaya, Eve M Valera, Harm Jan Van Der Horn, John D Van Horn, Ragini Verma, Benjamin S C Wade, Willian C Walker, Ashley L Ware, J Kent Werner, Keith Owen Yeates, Ross D Zafonte, Michael M Zeineh, Brandon Zielinski, Paul M Thompson, Frank G Hillary, David F Tate, Elisabeth A Wilde, Emily L Dennis

Faculty, Staff and Students Publications

An extensive library of symptom inventories has been developed over time to measure clinical symptoms of traumatic brain injury (TBI), but this variety has led to several long-standing issues. Most notably, results drawn from different settings and studies are not comparable. This creates a fundamental problem in TBI diagnostics and outcome prediction, namely that it is not possible to equate results drawn from distinct tools and symptom inventories. Here, we present an approach using semantic textual similarity (STS) to link symptoms and scores across previously incongruous symptom inventories by ranking item text similarities according to their conceptual likeness. We tested …

Skin Cancer Diagnosis Utilizing Hybrid Discrete Cosine Transform And High-Performance Convolutional Neural Networks, Mohammed M. Abo-Zahhad, Mohammed Abo-Zahhad

Mansoura Engineering Journal

Detecting skin cancer early and accurately is crucial for successfully treating this potentially fatal disease. Enhancing the accuracy of visual inspection techniques is often necessary to improve clinical decision-making and increase the chance of successful treatment outcomes. In this research, high-performance deep learning (DL) models for automated skin cancer categorization and early skin cancer diagnosis screening are developed and evaluated in conjunction with the discrete cosine transform (DCT). For this purpose, features are extracted from medical images using both techniques. The DCT is used for feature extraction and dimensionality reduction, while deep learning (DL) trains fully connected models for classification …

Quantum Chemical Methods And Multiscale Modeling In Computer-Aided Drug Design, Hunter W. La Force

Chemistry Theses and Dissertations

Computer-Aided Drug Design (CADD) leverages a diverse toolkit of computational methods to accelerate the discovery and development of novel therapeutics. Among these, quantum chemical calculations provide unparalleled accuracy in understanding molecular interactions, albeit at a higher computational cost. This accuracy is crucial for identifying and quantifying fundamental interactions that dictate drug efficacy and selectivity, as exemplified by our investigation of ruthenium polypyridyl complexes. These metal-based compounds are model systems for studying covalent coordination bonds between ruthenium and its ligands and noncovalent interactions with DNA and protein targets. Local Mode Vibrational Theory emerges as a powerful lens within this framework, enabling …

Terraincraft: Automated Land-Cover–Driven Terrain Generation For Marine Robot Simulations, Xinyue Liang

Dartmouth College Master’s Theses

From self-driving cars navigating city streets to all-terrain vehicles tackling rugged landscapes, recent leaps in robotic autonomy due to fast pace development in deep learning are reshaping how machines interact with the real world. However, autonomy in the aquatic environment is still limited, due to difficulty in testing and unavailability of realistic simulation environments.

In this project, we aim to create an automated system that simplifies the processes of creating synthetic datasets for marine robots navigation training tasks. We achieved this through a land cover map controlled terrain generation. Our goal is to provide an automatic terrain generation system that …

Machine Learning Course: A 15-Week Interactive Curriculum With Code And Case Studies, Pegah Khosravi

Open Educational Resources

This open-access machine learning course is a comprehensive 15-week curriculum developed and published on GitHub with full Google Colab compatibility. It combines theoretical concepts with hands-on Python coding, real-world datasets, and structured projects covering regression, classification, clustering, deep learning, transformers, and multimodal AI. The course is designed for students, educators, and researchers interested in applied machine learning, including biomedical applications. It includes explainable AI components and ethical discussions to align with modern AI standards. The course is maintained by BioMind AI Lab at CUNY.

A Comparison Of Surrogate Modeling Techniques For Predicting Radial Turbomachinery Flow Fields, Chase Oliphant

Utah Space Grant Consortium

Radial pumps and compressors are used in various engineering applications, including rocket turbopumps, automotive turbochargers, refrigeration systems, etc. Several different physical effects, including viscous losses, flow separation, compressibility, and rotational dynamics, dominate the flow in a radial impeller. These effects make predicting the flow field in a radial impeller very difficult, requiring computationally expensive CFD. However, a majority of an individual flow field calculated using CFD is often not pertinent to a designer, who only requires information at certain positions within the flow field. This often results in most of the simulation data generated being unused. Accurately predicting the flow …

Igtp: Learning Interpretable Cellular Embedding For Inferring Biological Mechanisms Underlying Single-Cell Transcriptomics, Kang-Lin Hsieh, Kai Zhang, Yan Chu, Lishan Yu, Xiaoyang Li, Nuo Hu, Isha Kawosa, Patrick G Pilié, Pratip K Bhattacharya, Degui Zhi, Xiaoqian Jiang, Zhongming Zhao, Yulin Dai

Faculty, Staff and Student Publications

Deep-learning models like Variational AutoEncoder have enabled low dimensional cellular embedding representation for large-scale single-cell transcriptomes and shown great flexibility in downstream tasks. However, biologically meaningful latent space is usually missing if no specific structure is designed. Here, we engineered a novel interpretable generative transcriptional program (iGTP) framework that could model the importance of transcriptional program (TP) space and protein-protein interactions (PPI) between different biological states. We demonstrated the performance of iGTP in a diverse biological context using gene ontology, canonical pathway, and different PPI curation. iGTP not only elucidated the ground truth of cellular responses but also surpassed other …

A Novel Approach To Attention-Based Models In Image Completion: Weighted Spatial-Attention Using Radial Distance, Tyler D. Kuper

Electrical Engineering and Computer Science Undergraduate Honors Theses

Humans infer missing visual information by focusing on spatial relationships in the context of their surroundings. Machine learning aims to replicate this skill through image completion, a fundamental task in current computer vision research. While advances in self-attention layers have recently enhanced generative machine learning models for text, these mechanisms still currently lack the capability to handle sparse image completion efficiently. We introduce a distance-based attention mechanism that uses radial-based weights to efficiently reconstruct an image. We compare this attention mechanism with self-attention and a fully connected network on an image completion task using the MNIST dataset. Our results show …

Implementation Of Residual Tandem Neural Networks For Photonic Inverse Design, Ponthea A. Zahraii

Electrical Engineering and Computer Science (MS) Theses

Deep-learning approaches can greatly benefit the modeling and design of nanophotonic and optical structures. Traditional full-wave simulations are time and resource-intensive, which can act as a bottleneck in photonic design. On the other hand, deep-learning approaches for designing the response of nanophotonic geometries can be computationally inexpensive and produce accurate and efficient results. In this project, we specifically investigate the case of optical forces near meta-structures. We propose using an inverse design approach with residual blocks to account for the deep nature of this architecture and inherently address the non-uniqueness problem. A tandem approach, which consists of two interconnected models, …

Toward Robust Semantic Segmentation In Levee Infrastructure Monitoring: Enhancing Accuracy With High-Fidelity Synthetic Data And Ensemble Learning, Padam Jung Thapa

LSU New Orleans Theses and Dissertations

Abstract: Levees serve as critical flood protection structures, but failures due to inadequate maintenance and extreme water pressures have led to devastating events such as Hurricane Katrina. Manual inspections are slow, labor-intensive, and prone to human error, necessitating the development of automated solutions. This study proposes an AI-driven framework for levee inspection utilizing deep learning-based semantic segmentation to detect rutting and enhance the identification of sand boils. To address dataset limitations, high-fidelity synthetic images are generated using DreamBooth for fine-tuning, while ControlNet adds structural constraints to enhance realism and consistency. A semi-automatic convex hull annotation technique enhances labeling efficiency, and …

Dose Prediction Via Deep Learning To Enhance Treatment Planning Of Lung Radiotherapy Including Simultaneous Integrated Boost Techniques, Wenhua Cao, Mary Gronberg, Stephen Bilton, Hana Baroudi, Skylar Gay, Christopher Peeler, Zhongxing Liao, Thomas J Whitaker, Karen Hoffman, Laurence E Court

Faculty, Staff and Student Publications

Background: Recent studies have shown deep learning techniques are able to predict three-dimensional (3D) dose distributions of radiotherapy treatment plans. However, their use in dose prediction for treatments with varied prescription doses including simultaneous integrated boost (SIB), that is, using multiple prescription doses within the same plan, and benefit in improving plan quality should be validated.

Purpose: To investigate the feasibility and potential benefit of using deep learning to predict dose distribution of volumetric modulated arc therapy (VMAT) including SIB techniques and improve treatment planning for patients with lung cancer.

Methods: The dose prediction model was trained with 93 retrospective …

T2-Weighted Imaging Of Rectal Cancer Using A 3d Fast Spin Echo Sequence With And Without Deep Learning Reconstruction: A Reader Study, Dan Nguyen, Sarah Palmquist, Ken-Pin Hwang, Jingfei Ma, Usama Salem, Jia Sun, Xinzeng Wang, Jong Bum Son, Randy Ernst, Peng Wei, Harmeet Kaur, Nir Stanietzky

Faculty, Staff and Student Publications

Purpose: To compare image quality and clinical utility of a T2-weighted (T2W) 3-dimensional (3D) fast spin echo (FSE) sequence using deep learning reconstruction (DLR) versus conventional reconstruction for rectal magnetic resonance imaging (MRI).

Methods: The study included 50 patients with rectal cancer who underwent rectal MRI consecutively between July 7, 2020 and January 20, 2021 using a T2W 3D FSE sequence with DLR and conventional reconstruction. Three radiologists reviewed the two sets of images, scoring overall SNR, motion artifacts, and overall image quality on a 3-point scale and indicating clinical preference for DLR or conventional reconstruction based on those three …

Few-Shot Learning On Graphs: From Meta-Learning To Llm-Empowered Pre-Training And Beyond, Yuan Fang, Yuxia Wu, Xingtong Yu, Shirui Pan

Research Collection School Of Computing and Information Systems

Graph representation learning has become central to many graph-based tasks, driving advancements in various domains such as web search, recommendation systems, and social network analysis. Traditionally, these methods rely on end-to-end supervised learning paradigms that require abundant labeled data, which can be costly and difficult to obtain. To address this limitation, few-shot learning on graphs has emerged as a promising approach, allowing models to generalize with minimal supervision and overcome data scarcity in real-world applications. This tutorial offers an in-depth exploration of recent advancements in few-shot learning for graphs, providing a comparative analysis of state-of-the-art methods and identifying future research …

Deep Learning Model With Hybrid Resampling Technique For Electricity Theft Detection In Smart Grid, Fawzia Algarni, Laila Nassef, Nahed Alowidi

Journal of King Abdulaziz University: Computing and Information Technology Sciences

Electricity theft is a serious problem that results in significant revenue losses and jeopardizes economic stability. The accurate detection of theft is hindered by the prevalence of imbalanced, real-world electricity consumption datasets. Consequently, data balancing is necessary prior to the classification of normal and abnormal consumers in order to mitigate the classifier’s inherent biases toward the majority class. This study proposes a deep learning–based electricity theft detection model coupled with a hybrid resampling technique (synthetic minority over-sampling technique–edited nearest neighbor (SMOTEENN)). This hybrid technique resolves the bias problem in classifying the imbalanced data, reduces the risk of overfitting, and preserves …

Deep Learning-Based Automatic Segmentation Of Cerebral Infarcts On Diffusion Mri, Wi-Sun Ryu, Dawid Schellingerhout, Jonghyeok Park, Jinyong Chung, Sang-Wuk Jeong, Dong-Seok Gwak, Beom Joon Kim, Joon-Tae Kim, Keun-Sik Hong, Kyung Bok Lee, Tai Hwan Park, Sang-Soon Park, Jong-Moo Park, Kyusik Kang, Yong-Jin Cho, Hong-Kyun Park, Byung-Chul Lee, Kyung-Ho Yu, Mi Sun Oh, Soo Joo Lee, Jae Guk Kim, Jae-Kwan Cha, Dae-Hyun Kim, Jun Lee, Man Seok Park, Dongmin Kim, Oh Young Bang, Eung Yeop Kim, Chul-Ho Sohn, Hosung Kim, Hee-Joon Bae, Dong-Eog Kim

Faculty, Staff and Student Publications

We explored effects of (1) training with various sample sizes of multi-site vs. single-site training data, (2) cross-site domain adaptation, and (3) data sources and features on the performance of algorithms segmenting cerebral infarcts on Magnetic Resonance Imaging (MRI). We used 10,820 annotated diffusion-weighted images (DWIs) from 10 university hospitals. Algorithms based on 3D U-net were trained using progressively larger subsamples (ranging from 217 to 8661), while internal testing employed a distinct set of 2159 DWIs. External validation was conducted using three unrelated datasets (n = 2777, 50, and 250). For domain adaptation, we utilized 50 to 1000 subsamples from …

Enhancing State-Of-The-Art Motor Imagery Classification With Reinforcement Learning, Anton Shepelev

USF Tampa Graduate Theses and Dissertations

One of the key obstacles to the rapid adoption of non-invasive Brain-Computer Interfaces (BCIs) for Motor Imagery (MI) is the low signal-to-noise ratio, and the substantial data requirements which can be mentally taxing for users. EEGNet, a compact Convolutional Neural Network (CNN), has long been considered the state-of-the-art (SOTA) for MI classification, demonstrating strong performance even with limited data. However, recent studies advocate for integrating Deep Reinforcement Learning (RL) to further enhance classification accuracy by dynamically optimizing feature extraction and decision-making processes. Despite this potential, practical implementations remain scarce due to challenges in stabilizing RL training and adapting it to …

Neural Network-Based Low-Level 3d Point Cloud Processing, Pingping Cai

Theses and Dissertations

3D computer vision is a promising research field with the potential to revolutionize future lifestyles. Among various 3D representation formats, point clouds stand out for their efficiency in depicting 3D objects using a set of coordinates, enabling advancements in fields such as autonomous driving, virtual reality, and robotics. Due to the limitations of sensor fields of view and scanning trajectories, the collected point clouds are usually sparse, noisy, and incomplete, impeding the performance of many downstream applications. Thus, the tasks of low-level point cloud processing are proposed to refine and generate dense, clean, and complete point clouds. To accomplish these …

Augmenting Deep Learning For Efficient Nextg Wireless Communication And Sensing Systems, Hem Kanta Regmi

Theses and Dissertations

Wireless networks have become an integral aspect of our daily lives. Over the years, earlier generations of wireless networks have enabled some innovative applications, such as wireless gaming, fast internet browsing, and home automation, which were previously unattainable. However, to support emerging technologies like autonomous driving, virtual reality, telemedicine, and intelligent manufacturing, which require high data throughput and low latency, there is a need for advanced wireless networks. Legacy networks like WiFi/LTE, which operate below 6 GHz, have limited bandwidth and are insufficient to fulfill the high data throughput demands of several applications. Millimeter-wave (mmWave) networks, operating between 30 GHz …

Applying Software Engineering Black-Box Methods For Testing Machine Learning Models, Timothy Elvira

Doctoral Dissertations and Master's Theses

This dissertation proposes researching an approach to incorporate and align Software black-box testing methods into Machine Learning (ML) applications, specifically in the context of computer vision models. Typically, testing methods within Software Engineering (SE) encompass a range of test types that assess levels of a software system, such as Unit, Integration, Functional, and System testing [1]. The testing spectrum offers two perspectives on the system: black-box, where the system’s code is hidden, and white-box, where the system's code is exposed for testing. Software Quality pairs testing with requirements, in a many-to-one relationship, to ensure proper validation of the software system. …

Optimized Deep Learning Model Using Binary Particle Swarm Optimization For Phishing Attack Detection: A Comparative Study, El-Sayed M. El-Kenawy, Marwa M. Eid, Hussein Lafta Hussein, Ahmed M. Osman, Ahmed M. Osman

Mesopotamian Journal of CyberSecurity

Phishing attacks manipulate users to disclose critical information, resulting in cybersecurity risks. Traditional phishing detection algorithms usually have large false positive rates and poor feature selection, degrading performance. This paper presents an optimized phishing detection framework that integrates binary particle swarm optimization (BPSO)-based feature selection (FS) with deep learning models. Six deep learning architectures were evaluated on the selected feature subset to identify the most effective model for accurate phishing classification. BPSO was used to select suitable attributes on a public Kaggle dataset with 10,000 samples, comprising phishing and legitimate website data with 48 attributes. NumDots, UrlLength, IpAddress, and NoHttps …

Ensure Privacy-Preserving Using Deep Learning, Abeer Dawood Salman, Ruqayah Rabeea Al-Dahhan

Mesopotamian Journal of CyberSecurity

Deep learning has emerged as a powerful approach for treating complex real-world challenges. However, the performance of the deep learning models is heavily reliant on access to large volumes of high-quality training data—an aspect often constrained by privacy concerns. Ensuring data availability while preserving user confidentiality remains a pressing issue. In response, cryptographic techniques like homomorphic encryption (HE), which are grounded in strict mathematical principles, present hopeful solutions for securing data on digital platforms without compromising its usability for learning models. It performs computations on encrypted data without revealing the underlying plaintext. The main attractive feature of this technique is …

Exploring Transfer Learning And Convolutional Autoencoder For Effective Kitchen Utensils Classification, Hashim Rosli, Rozniza Ali, Muhamad Suzuri Hitam, Ashanira Mat Deris, Noor Hafhizah Abd Rahim

Malaysian Journal of Computing (MJoC)

Effective classification of kitchen utensils is crucial for advancing assistive technologies and enhancing daily living for individuals with visual impairments. This study investigates the use of transfer learning and convolutional autoencoders to improve classification accuracy. We integrate pre-trained networks into an autoencoder framework to enhance feature extraction and image reconstruction. Models including ResNet50, DenseNet121, and their autoencoder variants were evaluated using precision, recall, accuracy, Structural Similarity Index Measure (SSIM), and Peak Signal-to-Noise Ratio (PSNR). Results show that DenseNet121 outperforms ResNet50 with a classification accuracy of 72% and shorter training time. When combined with autoencoders, DenseNet121-Autoencoder achieves the highest classification accuracy …

Examining The Impact Of Feature Selection Techniques On Machine And Deep Learning Models For The Prediction Of Covid 19, Hafiza Zoya Mojahid, Jasni Mohamad Zain, Marina Yusoff, Abdul Basit, Abdul Kadir Jumaat, Mushtaq Ali

Malaysian Journal of Computing (MJoC)

Feature selection is a vital preprocessing step for identifying the most informative features in complex datasets, enhancing the efficiency and accuracy of machine learning models. Its applications extend across various domains, including big data analytics, finance, chemometrics, medical diagnostics, biological research, intrusion detection systems, and renewable energy solutions. In medical contexts, feature selection serves a dual purpose: it reduces dimensionality while simultaneously improving the comprehension of disease etiology. This study delves into key variable selection methods—specifically Recursive Feature Elimination (RFE), Principal Component Analysis (PCA) and Least Absolute Shrinkage and Selection Operator (LASSO). We evaluate the interaction of these methods with …

Multimodal Ai-Driven Biomarker For Early Detection Of Cancer Cachexia, Sabeen Ahmed

USF Tampa Graduate Theses and Dissertations

Cancer cachexia is a metabolic syndrome characterized by substantial skeletal muscle loss, impacting cancer patients' survival and quality of life. Despite its clinical significance, early detection remains a challenge due to the lack of standardized diagnostic criteria and the reliance on indirect markers. This work presents an AI-driven approach to enhance cachexia detection and monitoring by integrating multiple deep learning methodologies. We explore transformer architectures for time-series analysis to model sequential medical data, enabling disease prediction and progression modeling. To ensure robust and reliable decision-making in clinical settings, we explore Bayesian deep neural networks for uncertainty estimation. Additionally, we introduce …

Leveraging Large Language Models For Knowledge-Free Weak Supervision In Clinical Natural Language Processing, Enshuo Hsu, Kirk Roberts

Faculty, Staff and Student Publications

The performance of deep learning-based natural language processing systems is based on large amounts of labeled training data which, in the clinical domain, are not easily available or affordable. Weak supervision and in-context learning offer partial solutions to this issue, particularly using large language models (LLMs), but their performance still trails traditional supervised methods with moderate amounts of gold-standard data. In particular, inferencing with LLMs is computationally heavy. We propose an approach leveraging fine-tuning LLMs and weak supervision with virtually no domain knowledge that still achieves consistently dominant performance. Using a prompt-based approach, the LLM is used to generate weakly-labeled …