Computer Sciences Commons^™

Speculative Automated Refactoring Of Imperative Deep Learning Programs To Graph Execution, Raffi Khatchadourian Ph.D., Tatiana Castro Vélez, Mehdi Bagherzadeh, Nan Jia, Anita Raja

Publications and Research

Efficiency is essential to support ever-growing datasets, especially for Deep Learning (DL) systems. DL frameworks have traditionally embraced deferred execution-style DL code---supporting symbolic, graph-based Deep Neural Network (DNN) computation. While scalable, such development is error-prone, non-intuitive, and difficult to debug. Consequently, more natural, imperative DL frameworks encouraging eager execution have emerged but at the expense of run-time performance. Though hybrid approaches aim for the "best of both worlds," using them effectively requires subtle considerations. Our key insight is that, while DL programs typically execute sequentially, hybridizing imperative DL code resembles parallelizing sequential code in traditional systems. Inspired by this, we …

Speculative Automated Refactoring Of Imperative Deep Learning Programs To Graph Execution, Raffi Khatchadourian Ph.D., Tatiana Castro Vélez, Mehdi Bagherzadeh, Nan Jia, Anita Raja

Publications and Research

Efficiency is essential to support ever-growing datasets, especially for Deep Learning (DL) systems. DL frameworks have traditionally embraced deferred execution-style DL code---supporting symbolic, graph-based Deep Neural Network (DNN) computation. While scalable, such development is error-prone, non-intuitive, and difficult to debug. Consequently, more natural, imperative DL frameworks encouraging eager execution have emerged but at the expense of run-time performance. Though hybrid approaches aim for the "best of both worlds," using them effectively requires subtle considerations. Our key insight is that, while DL programs typically execute sequentially, hybridizing imperative DL code resembles parallelizing sequential code in traditional systems. Inspired by this, we …

Speculative Automated Refactoring Of Imperative Deep Learning Programs To Graph Execution, Raffi T. Khatchadourian Ph.D., Tatiana Castro Vélez, Mehdi Bagherzadeh, Nan Jia, Anita Raja

Publications and Research

Efficiency is essential to support ever-growing datasets, especially for Deep Learning (DL) systems. DL frameworks have traditionally embraced deferred execution-style DL code---supporting symbolic, graph-based Deep Neural Network (DNN) computation. While scalable, such development is error-prone, non-intuitive, and difficult to debug. Consequently, more natural, imperative DL frameworks encouraging eager execution have emerged but at the expense of run-time performance. Though hybrid approaches aim for the "best of both worlds," using them effectively requires subtle considerations. Our key insight is that, while DL programs typically execute sequentially, hybridizing imperative DL code resembles parallelizing sequential code in traditional systems. Inspired by this, we …

When Deep Learning Meets Information Retrieval-Based Bug Localization: A Survey, Feifei Niu, Chuanyi Li, Kui Liu, Xin Xia, David Lo

Research Collection School Of Computing and Information Systems

Bug localization is a crucial aspect of software maintenance, running through the entire software lifecycle. Information retrieval-based bug localization (IRBL) identifies buggy code based on bug reports, expediting the bug resolution process for developers. Recent years have witnessed significant achievements in IRBL, propelled by the widespread adoption of deep learning (DL). To provide a comprehensive overview of the current state of the art and delve into key issues, we conduct a survey encompassing 61 IRBL studies leveraging DL. We summarize best practices in each phase of the IRBL workflow, undertake a meta-analysis of prior studies, and suggest future research directions. …

Image Captioning Through The Lens Of The Gricean Maxims: Generating Meaningful And Relevant Image Descriptions, Annika Lindh

Doctoral

Image captioning models enable us to automatically generate natural language image descriptions for previously unseen images. It combines the two fields of computer vision and natural language generation, allowing models to interpret the con tent of an image and communicate that knowledge through natural language text.

Research into image captioning has the potential benefit of reducing the gap in digital information availability between fully sighted individuals and those who are visually impaired. However, automatically generated captions often fail to provide the required level of detail and specificity to achieve this goal. Furthermore, current standard evaluation methods are insufficient at measuring …

Damslnet: Dual-Attention Multi-Scale Lightweight Network For Plant Disease Classification, Linfan Deng, Juan Qin, Kun Li, Jinhua Zhu, Zhaoxia Wang

Research Collection School Of Computing and Information Systems

Accurately identifying crop diseases plays a crucial role in advancing intelligent and modern agricultural production. Deep learning techniques have performed robust performance in classifying plant disease images. However, current studies face the challenge that many plant disease datasets are generated in controlled environments, leading to reduced model performance in real-world agricultural settings. This paper aims to provide a lightweight model that can accurately classify plant diseases in natural environments. Specifically, this paper investigates the Dual-Attention Multi-Scale Lightweight Network (DAMSLNet), which combines dual-attention-based multi-scale feature extraction and deep information fusion, to classify plant diseases. At the front end, the model employs …

Deep Learning Based Contactless Fingerprint Identification, Mohammad Alsmirat, M. Moneb Khaled, Aghyad A.L. Sayadi

Faculty Publications

Biometric authentication systems, particularly contactless fingerprint methods, offer enhanced security and convenience across various domains like access control, law enforcement, and finance. Despite these advantages, contactless systems face significant challenges related to image quality, finger orientation, and environmental factors. To address this, our paper presents the first extensive deep learning-based study on contactless fingerprint recognition using a large dataset of 2,143 images from 175 individuals. Our proposed approach integrates state-of-the-art preprocessing techniques with deep learning models to boost identification performance. After studying various transfer learning models, we achieved a high accuracy of 93.5%. We also conducted two further studies on …

Soft Sensor Modeling Based On Improved Transformer In Dual-Stream Framework, Hao Gu, Jiayu Wang, Weili Xiong

Journal of System Simulation

Abstract: Industrial process information is highly nonlinear and dynamic, with long-term dependencies between data, making it difficult to adequately extract time-series features. To address this issue, an improved Transformer-based soft sensor model in a dual-stream framework was proposed. The data were segmented and expanded. The features were extracted in parallel using a dual-stream structure combining a convolutional neural network with a self-attention mechanism and the improved Transformer model. The dual-stream features were fused for soft sensor regression. Residual connections were further introduced to accelerate the convergence speed of the model, and an orthogonal random features-based improved multi-head attention mechanism was …

Deep Learning For Hate Speech Detection: A Comparative Study, Jitendra Singh Malik, Hezhe Qiao, Guansong Pang, Anton Van Den Hengel

Research Collection School Of Computing and Information Systems

Automated hate speech detection is an important tool in combating the spread of hate speech, particularly in social media. Numerous methods have been developed for the task, including a recent proliferation of deep-learning based approaches. A variety of datasets have also been developed, exemplifying various manifestations of the hate-speech detection problem. We present here a largescale empirical comparison of deep and shallow hate-speech detection methods, mediated through the three most commonly used datasets. Our goal is to illuminate progress in the area, and identify strengths and weaknesses in the current state-of-the-art. We particularly focus our analysis on measures of practical …

Artificial Intelligence In Head And Neck Cancer: Towards Precision Medicine, Jacob Hagen, Logan Hornung, William Barham, Supratik Mukhopadhyay, Adam Bess, Kevin Contrera, Devraj Basu, Vlad Sandulache, Guillaume Spielmann, Sagar Kansara

School of Medicine Faculty Publications

Over the past 20 years, the capabilities of artificial intelligence (AI) have gained significant interest. While AI has been implemented to various degrees in several disciplines, its unique applications in head and neck cancer (HNC) remain underdeveloped. This narrative review examines the existing body of literature regarding the use of AI in HNC. Studies to date have demonstrated AI’s utility across multiple phases of the HNC treatment continuum. Despite its promise, integrating AI into clinical practice faces several challenges, including concerns about system integrity, generalizability, privacy, and bias. In this review, we address these challenges and offer insights into future …

Sugar: A Sequence Unfolding Based Transformer Model For Group Activity Recognition, Yash U. Gondkar

Graduate Masters Theses

Large Language Models have improved significantly in the past couple of years due to the adoption of transformers. However, transformers still find it challenging to process videos due to limited context size caused by their quadratic computing cost. Therefore, we studied a booming field in machine learning which powers applications like social scene analysis and video surveillance systems called Group Activity Recognition (GAR). We found that recent models were able to achieve more than 90% accuracy on popular datasets like the Volleyball dataset, however, it turned out that even they relied on transformers.

Therefore, in this work, we developed a …

Ddos Attack Detection In Edge-Iiot Digital Twin Environment Using Deep Learning Approach, Feras Al-Obeidat, Adnan Amin, Ahmed Shuhaiber, Inam Ul Haq

All Works

The industrial Internet of Things (IIoT) and digital twins are redefining how digital models and physical systems interact. IIoT connects physical intelligence, and digital twins virtually represent their physical counterparts. With the rapid growth of Edge-IIoT, it is crucial to create security and privacy regulations to prevent vulnerabilities and threats (i.e., distributed denial of service (DDoS)). DDoS attacks use botnets to overload the target system with requests. In this study, we introduce a novel approach for detecting DDoS attacks in an Edge-IIoT digital twin-based generated dataset. The proposed approach is designed to retain already learned knowledge and easily adapt to …

Domain Obedient Deep Learning, Soumadeep Saha

Doctoral Theses

Deep learning, a family of data-driven artificial intelligence techniques, has shown immense promise in a plethora of applications, and it has even outpaced experts in several domains. However, unlike symbolic approaches to learning, these methods fall short when it comes to abiding by and learning from pre-existing established principles. This is a significant deficit for deployment in critical applications such as robotics, medicine, industrial automation, etc. For a decision system to be considered for adoption in such fields, it must demonstrate the ability to adhere to specified constraints, an ability missing in deep learning-based approaches. Exploring this problem serves as …

Predicting Music Origin With Deep Learning, Fruzsina Ladanyi

Electronic Theses, Projects, and Dissertations

This project explores the usage of a late fusion deep learning architecture to predict the geographic origin of music. Mel-Frequency Cepstral Coefficients (MFCCs) and the language of the music sample are used as features. MFCCs were extracted from audio files to capture sound features. The language was identified using OpenAI’s Whisper model to provide additional context. A late fusion neural network architecture combining Long Short-Term Memory (LSTM) layers for sequential MFCC input and dense layers for non-sequential language features were employed to support both classification and regression tasks. The classification model achieved an accuracy of 33.03% across 56 countries or …

Roadside Asset Extraction From Mobile Lidar Point Cloud, Yushin Ahn, Riadh Munjy, Stephen Choi

Mineta Transportation Institute

Mobile LiDAR systems are powerful tools that help us map roads and their surroundings in 3D with great speed and precision. The data provided by these systems support urban planning efforts, digital mapping, transportation infrastructure maintenance, and more. This report presents a comprehensive workflow for roadside asset extraction using Mobile Terrestrial Laser Scanning (MTLS) data, focusing on road lane detection, cross-section slope analysis, and point cloud classification. Roadside asset extraction is the identification and classification of roadside features like signs and poles. The dataset, acquired using a high-resolution mobile LiDAR system, contains over 5.7 billion points (pieces of data) across …

Use Of A Preliminary Artificial Intelligence-Based Laryngeal Cancer Screening Framework For Low-Resource Settings: Development And Validation Study, Shao Wei Sean Lam, Min Hun Lee, Michael Dorosan, Samuel Altonji, Hiang Khoon Tan, Walter T. Lee

Research Collection School Of Computing and Information Systems

Background: Early-stage diagnosis of laryngeal cancer significantly improves patient survival and quality of life. However, the scarcity of specialists in low-resource settings hinders the timely review of flexible nasopharyngoscopy (FNS) videos, which are essential for accurate triage of at-risk patients.Objective: We introduce a preliminary AI-based screening framework to address this challenge for the triaging of at-risk patients in low-resource settings. This formative research addresses multiple challenges common in high-dimensional FNS videos: (1) selecting clear, informative images; (2) deriving regions within frames that show an anatomical landmark of interest; and (3) classifying patients into referral grades based on the FNS video …

Gene Regulatory Network Prediction Using Machine Learning, Deep Learning, And Hybrid Approaches, Sai Teja Mummadi, Md Khairul Islam, Victor Busov, Hairong Wei

Michigan Tech Publications

Construction of gene regulatory networks (GRNs) is essential for elucidating the regulatory mechanisms underlying metabolic pathways, biological processes, and complex traits. In this study, we developed and evaluated machine learning, deep learning, and hybrid approaches for constructing GRNs by integrating prior knowledge and large-scale transcriptomic data from Arabidopsis thaliana, poplar, and maize. Among these, hybrid models that combined convolutional neural networks and machine learning consistently outperformed traditional machine learning and statistical methods, achieving over 95% accuracy on the holdout test datasets. These models not only identified a greater number of known transcription factors regulating the lignin biosynthesis pathway but also …

Research On Requirements And Methods For Intelligent Assessment Of Simulation Credibility, Bingheng Wang, Tingrui Liu, Fan Yang, Huan Zhang, Wei Li, Ping Ma, Ming Yang

Journal of System Simulation

Abstract: The accuracy of simulations in representing real-world systems is a critical concern for users. Simulation credibility assessment ensures trustworthiness by evaluating the correctness and effectiveness of simulations to meet application requirements. As simulation technologies are widely adopted, and new simulation paradigms emerge, traditional assessment methods are increasingly showing limitations in their dependence on experts, data processing capabilities, and assessment efficiency. This paper systematically reviewed the research demands, current progress, new technologies, and future trends of intelligent simulation credibility assessment. Based on the simulation credibility assessment process and problem analysis, the requirements for intelligent credibility assessment were discussed. Intelligent technologies …

Machine Learning Crime Prediction Models And The Gap Between Research And Implementation: A Systematic Review, Ricardo Huamantingo, Miguel Cano-Lengua, Ciro Rodriguez

Karbala International Journal of Modern Science

A crime is an illegal or violent act committed by one individual against another. The increasing crime rate has become a major concern as it negatively affects people's quality of life and generates significant social and economic costs. This study aims to identify the most widely used machine learning (ML) models for crime prediction, determine evaluation metrics for assessing model performance, and analyze key data characteristics to enhance real-world implementation. The study follows the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) methodology. A search string was formulated using the population, intervention, comparison, and outcomes (PICO) framework and applied …

Graph Traverse Reference Network For Sign Language Corpus Retrieval In The Wild, Kun Hu, Fengxiang He, Adam Schembri, Zhiyong Wang

Research outputs 2022 to 2026

Sign languages are the primary languages of the deaf community as well as hearing individuals who are unable to speak, which engage the visual-manual modality to convey meanings. In recent years, there has been an explosive growth of sign language videos available from video streaming and social media service platforms. Given the size of these corpora, sign language users often face significant challenges in effectively acquiring the information they need. Therefore, we propose a novel deep learning architecture, namely Graph Traverse Reference Network (GTRN), allowing visual signing queries to retrieve relevant sign language videos (documents) from a large corpus. GTRN …

Adapting A Segmentation Foundation Model For Medical Image Classification, Pengfei Gu, Haoteng Tang, Islam A. Ebeid, Jose Angel Nunez, Fabian Vazquez Jr., Diego Adame

Computer Science Faculty Publications

Recent advancements in foundation models, such as the Segment Anything Model (SAM), have shown strong performance in various vision tasks, particularly image segmentation, due to their impressive zero-shot segmentation capabilities. However, effectively adapting such models for medical image classification is still a less explored topic. In this paper, we introduce a new framework to adapt SAM for medical image classification. First, we utilize the SAM image encoder as a feature extractor to capture segmentation-based features that convey important spatial and contextual details of the image, while freezing its weights to avoid unnecessary overhead during training. Next, we propose a novel …

Aircraft Bird Strike Risk Prediction Using Machine Learning And Analytic Hierarchy Process, Jason Anthony Powell

Doctoral Dissertations and Master's Theses

To address the limitations of Next Generation Radar-based bird strike forecasting, this study modeled 12 spatiotemporal weather features from the National Oceanic and Atmospheric Administration alongside bird strike risk using Long Short-Term Memory Recurrent Neural Network (LSTM-RNN), XGBoost regression tree, and Bayesian network algorithms. Five years of bird strike data from four geographically diverse airfields served as the target risk variable, categorized as low, moderate, or severe based on Department of the Air Force risk models. The ensemble model, which combines the LSTM-RNN and XGBoost regression algorithms, yielded the most accurate forecasts, achieving 80% to 93% accuracy across all airfields, …

Causal Neuro-Symbolic Artificial Intelligence: Synergy Between Neuro-Symbolic And Causal Artificial Intelligence, Utkarshani Jaimini

Theses and Dissertations

Understanding and reasoning about cause and effect is innate to human cognition. In everyday life, humans continuously engage in causal reasoning and hypothetical retrospection to make decisions, plan actions, and interpret events. This cognitive ability allows us to ask questions such as: “What caused this situation?”, “What will happen if I take this action?”, or “What would have happened had I chosen differently?” This intuitive capacity to form mental models of the world, infer causal relationships, and reason about alternative scenarios, particularly counterfactuals, is central to our intelligence and adaptability. In contrast, current machine learning (ML) and artificial intelligence (AI) …

Toward A Simplified Framework For Sequential Character Recognition, Nicholas Howe

Computer Science: Faculty Publications

This paper proposes a novel approach to handwritten charac- ter recognition using convolutional non-recurrent deep neural networks. Such a network can run in parallel at every point of a document, offer- ing potential advantages in speed over recurrent approaches. The net- work’s output feeds into a beam search optimization for final decoding. Preliminary quantitative results show that the framework can achieve bootstrap training from labeled word images. It provides an alternative to sequential models that rely on connectionist temporal classification for alignment.

A Comparative Study Of Machine Learning And Deep Learning Models In Binary And Multiclass Classification For Intrusion Detection Systems, Ayesha Alharthi, Meera Alaryani, Sanaa Kaddoura

All Works

Network infrastructure evolution has significantly expanded the attack surface, leading to increasingly complex and sophisticated cybersecurity threats. Traditional rule-based intrusion detection systems (IDS) often fail to detect emerging attack vectors, prompting the need for intelligent, data-driven approaches. This study evaluates and compares the performance of machine learning (ML) and deep learning (DL) models for network intrusion detection. Two publicly available datasets were utilized: a binary-labeled software-defined networking (SDN) dataset and a multiclass industrial control system dataset based on the IEC 60870-5-104 protocol. Preprocessing steps included normalization, label encoding, and a 70:10:20 train-validation-test split. Seven models, Random Forest, Decision Tree, K-Nearest …

Simulation Study On Adaptive Signal Control Of Deformed Intersection Based On Lstm-Gnn, Kun Chen, Liang Chen, Jiming Xie, Fengbo Liu, Taixiong Chen, Lukuan Wei

Journal of System Simulation

Abstract: Aiming at the traffic congestion at deformed intersections, an improved adaptive traffic signal control scheme based on deep learning is designed, the scheme integrates the adaptive signal control of LSTM and GNN at deformed intersections. LSTM is used to capture the dependence between time series traffic data, while GNN is used to construct a spatial interaction model between lanes. By integrating the information of time and space dimensions, the model can dynamically adjust the phase duration of signal lights according to real-time traffic conditions. The results indicate that the LSTM-GNN adaptive control scheme improves overall traffic throughput efficiency by …

Arson Event Detection Using Yolov9, Ali Abbas Abbod, Matheel E. Abdulmunimb, Ismail A. Mageed

Journal of Soft Computing and Computer Applications

Detecting event anomalies is crucial for surveillance systems, as it enables the identification of occurrences in videos, both temporally and spatially. It can identify deviations from patterns without requiring human oversight by learning from past information to distinguish normal behavior and pinpoint irregularities. Early detection of arson fires is critical to mitigating damage, public safety, property, and the environment, as well as saving lives and aiding in law enforcement investigations. The objective of this study is to evaluate a system for detecting events using the You Only Look Once version 9 (YOLOv9) model in surveillance videos with a focus on …

Application Of Deep Learning In Analysis Of Stellar Spectra, Piyush Yayati

Master’s Dissertations

In recent years, the analysis of high-resolution stellar spectra has become increasingly important for estimating key stellar parameters such as effective temperature (Teff ), surface gravity (log g), metallicity ([M/H]), and rotational velocity (v sin i). Traditional methods often rely on manual calibration or spectrum synthesis, which can be time-consuming and error-prone, especially for M dwarfs whose spectra are dense with molecular features. In this study, we investigate the use of convolutional neural networks (CNNs) to automate the estimation of stellar parameters using synthetic and observed data.We adopt a StarNet-like CNN architecture trained on synthetic spectra generated from the PHOENIX-ACES …

Enabling Automatic Solar Pv Array Identification Using Big Satellite Imagery, Qi Li, Keyang Yu, Carson Snow, Dong Chen

Computer Science Faculty Research and Publications

Recently, there has been a growing interest in automatically collecting distributed solar photovoltaic (PV) installation information in smart grid systems, including the quantity and locations of solar PV deployments, as well as their profiling information across a given geospatial region. Most recent approaches are still suffering low detection accuracy due to insufficient sample and principal feature learning when building their models and also separation of rooftop object segmentation and identification during their detection processes. In addition, they cannot report accurate multi-deployment results. To address these problems, we design a new system-SolarDetector+, which can automatically and accurately detect and profile distributed …

Advancing Fake News Detection With Graph Neural Network And Deep Learning, Haji Gul, Feras Al-Obeidat, Muhammad Wasim, Adnan Amin, Fernando Moreira

All Works

In the modern era of digital technology, the rapid distribution of news via social media platforms substantially contributes to the propagation of false information, presenting challenges in upholding the accuracy and reliability of information. This study presents an updated approach that utilizes graph neural networks (GNNs) alongside with advanced deep learning techniques to improve the identification of false information. In contrast to traditional approaches that primarily rely on analyzing text and assessing the credibility of sources, our methodology utilizes the structural information of news propagation networks. This allows for a detailed comprehension of the interconnections and patterns that are indicative …