Computer Sciences Commons^™

From Release To Adoption: Challenges In Reusing Pre-Trained Ai Models For Downstream Developers, Peerachai Banyongrakkul, Mansooreh Zahedi, Patanamon Thongtanunam, Christoph Treude, Haoyu Gao

Research Collection School Of Computing and Information Systems

Pre-trained models (PTMs) have gained widespread popularity and achieved remarkable success across various fields, driven by their groundbreaking performance and easy accessibility through hosting providers. However, the challenges faced by downstream developers in reusing PTMs in software systems are less explored. To bridge this knowledge gap, we qualitatively created and analyzed a dataset of 840 PTM-related issue reports from 31 OSS GitHub projects. We systematically developed a comprehensive taxonomy of PTM-related challenges that developers face in downstream projects. Our study identifies seven key categories of challenges that downstream developers face in reusing PTMs, such as model usage, model performance, and …

Large Lithium-Ion Battery Model For Secure Shared E-Bike Battery In Smart Cities, Donghui Ding, Zhao Li, Linhao Luo, Ming Jin, Bin Zhu, Yichen Zhong, Junhao Hu, Peng Cai, Huiqi Hu

Research Collection School Of Computing and Information Systems

Electric bikes powered by lithium-ion batteries are increasingly used in smart cities to promote sustainable mobility and efficient delivery services. However, limited battery range and slow plug-in charging remain key challenges. Shared electric bike battery systems, facilitated by battery swapping stations, offer a promising solution by enabling quick and efficient battery replacements. However, their success hinges on accurate anomaly detection, battery health estimation and remain range prediction. These tasks remain challenging due to data scarcity, battery diversity and environmental variability. Here we show that a large-scale lithium-ion battery model trained on over ten million battery time series data enables robust …

Stylegan-∞: Extending Stylegan To Arbitrary-Ratio Translation With Stylebook, Yihua Dai, Tianyi Xiang, Bailin Deng, Yong Du, Hongmin Cai, Jing Qin, Shengfeng He

Research Collection School Of Computing and Information Systems

Although pre-trained large-scale generative models StyleGAN series have proven to be effective in various editing and translation tasks, they are limited to pre-defined fixed aspect ratio. To overcome this limitation, we propose StyleGAN-∞, a model that enables pre-trained StyleGAN to perform arbitrary-ratio conditional synthesis. Our key insight is to distill the expressive StyleGAN features into a StyleBook, such that an arbitrary-ratio condition can be translated to other forms by properly assembling pre-defined StyleBook vectors. To learn and leverage the StyleBook, we employ a network with three distinct stages, each corresponding to StyleBook extraction, StyleBook correspondence learning, and arbitrary-ratio synthesis. Extensive …

Multi-Fidelity Machine Learning Modeling For Aerodynamic Response Prediction Of Aerospace Vehicles, Ethan S. Jackman

Theses and Dissertations

Hypersonic vehicle design requires understanding complex aerodynamic phenomena across the full flight regime. This study presents a novel MF surrogate modeling methodology that enables the prediction the full field response across a vehicle’s surface. A Space-Filling Curve (SFC) is used to convert unstructured data into 1D vectors. The a Convolutional Autoencoder is used with transfer learning to reduce the dimensionality of the data. An Emulator-Embedded Neural Network (E2NN) combines multi-fidelity data for fast, accurate predictions. A benchmark analytical example and hypersonic application are used to evaluate the methodology. Using various numbers of samples and sampling strategies it is found that …

Optimizing Fire Detection In Remote Sensing Imagery For Edge Devices: A Quantization-Enhanced Hybrid Deep Learning Model, Syed Muhammad Salman Bukhari, Nadia Dahmani, Sujan Gyawali, Muhammad Hamza Zafar, Filippo Sanfilippo, Kiran Raja

All Works

Wildfires are increasing in frequency and severity, presenting critical challenges for timely detection and response, particularly in remote or resource-limited environments. This study introduces the Inception-ResNet Transformer with Quantization (IRTQ), a novel hybrid deep learning (DL) framework that integrates multi-scale feature extraction with global attention and advanced quantization. The proposed model is specifically optimized for edge deployment on platforms such as unmanned aerial vehicles (UAVs), offering a unique combination of high accuracy, low latency, and compact memory footprint. The IRTQ model achieves 98.9% accuracy across diverse datasets and shows strong generalization through cross-dataset validation. Quantization significantly reduces the parameter count …

Futurescape Libraries Ai Toolkit, Keith Webster

Copyright, Fair Use, Scholarly Communication, etc.

A toolkit developed to explore scenario-specific strategies and activities that research libraries can undertake to prepare for various possible AI-influenced futures. The toolkit integrates the ARL/CNI AI Scenarios published in spring 2024 along with priorities trialed and refined by strategic thinkers working directly in, or adjacent to, the research library field during a Strategic Implications forum held December 7–8, 2024, in Washington, DC.

A Dynamic Hierarchical Attention Framework For Multimodal Malware Detection, Tamanna Nazmin

Graduate Theses and Dissertations

The increasing use of Android in the worldwide mobile ecosystem has come along with a significant increase in advanced malware, highlighting the critical necessity for efficient, scalable, and adaptable detection systems. Despite recent advancements in machine learning improving malware detection, the majority of current solutions are limited to one, two, or three data modalities, hence neglecting the comprehensive behavioral spectrum of contemporary multi-vector threats. This thesis presents the first comprehensive multimodal framework for Android malware detection, which combines textual, time-series (temporal), graph-based (structural), and visual information using an innovative hierarchical attention mechanism and Dynamic Fusion Controller(DFC). Our methodology consistently classifies …

Script-Based Inferences In An Image Schema Story Understander, Jamie C. Macbeth, Boming Zhang, Sharmin Badhan

Computer Science: Faculty Publications

Recent studies of large language models (LLMs) have revealed that they lack human-like cognitive models of reasoning and understanding. An important thread of research merges image schemas into symbolic artificial intelligence systems where their use as conceptual building blocks and primitives shows promise for the study of human cognition through intelligent systems that perform neurosymbolically. The work presented in this paper demonstrates image schema primitives being used in structures of a representation system called conceptual dependency (CD) and in broader commonsense knowledge structures called scripts. We present a story under- standing system that uses image schemas as primitives in scripts …

Applying Artificial Intelligence And Chatbots To Enhance Collection Development In Health Sciences Libraries, Ivan Portillo, David Carson

Library Presentations, Posters, and Audiovisual Materials

The continuous advancement of artificial intelligence (AI) and large language models (LLMs) has presented several opportunities for librarians to reduce their workload and become more efficient. This session will explore the potential of generative AI chatbots in assisting health sciences librarians with collection development. Two methods that will be discussed include the potential of AI to help discover new titles and how AI can evaluate your library collection for any potential gaps based on a college program’s curriculum.

Multi-Modal Depth Estimation Using Camera And Mmwave Sensors, Alston D. Devero-Belfon

Student Theses

For accurately estimating the depth of environments with varying lighting conditions, reliable methods are limited. By utilizing wireless sensor technology in conjunction with cameras, a wide range of environments can be visualized, and objects within these environments can be tracked and monitored. Such methods offer cost-effective alternatives and provide a more secure, data-at-rest option for individuals with low vision, while also enhancing machine perception. In this work, we develop such a prototype that utilizes wireless sensors and cameras, which act in sync, enabling us to estimate the depth of objects within varying lighting environments to a level that is recognizable …

Ibi-Dt: A Novel Approach Combining Individualized Bayesian Inference And Decision Tree For Identifying Cancer Drivers And Their Interactions, Md Asad Rahman, Gregory F. Cooper, Jinying Zhao, Xinghua Lu, Jinling Liu

Engineering Management and Systems Engineering Faculty Research & Creative Works

Cancer is mainly caused by a relatively small portion of somatic genome alterations (SGAs), called cancer drivers. Despite success in identifying a good number of cancer drivers, many more remain to be discovered to explain various cancers. Moreover, limited tools are available to identify potential interactions among cancer drivers for a better understanding of oncogenesis. To tackle these challenges, we have developed a novel approach called individualized Bayesian inference using a decision tree (IBI-DT). IBI-DT recognizes the genetic heterogeneity among cancer patients, where different individuals or patient subgroups of distinct genomic makeup may have different drivers. IBI-DT works by constructing …

The “Founder’S Gaze”: How The Fourth Amendment Is A Surveillance Technology That Enables Ai To Scale Control Over The Subaltern, Diego H. Alcalá Laboy

Michigan Journal of Race and Law

Much has been written about the rise of artificial intelligence and machine learning applications and how the current Fourth Amendment law has been unable to mitigate the privacy harm that these tools produce. This article explores how the development and usage of AI and machine learning models is dependent on the originalism principles of Fourth Amendment Law. Utilizing Critical Surveillance Studies and Anticolonial Theory, I posit that the Fourth Amendment is a surveillance technology that categorizes conduct, persons, and places to impose the material conditions for the subjugation of historically minoritized communities within the United States. Furthermore, this article explores …

The Density Distribution Of The Material Around Betelgeuse, Iman Behbehani

Dissertations, Theses, and Capstone Projects

Betelgeuse is a red supergiant star visible in the constellation Orion. Its windy and highly convective surface results in a complicated mass loss pattern difficult to understand and replicate in simulations. The ejected mass can form a shell around the star we consider the circumstellar material (CSM). In this study, we use ALMA interferometric observations to find the structure of Betelgeuse's CSM, and connect to the mass loss mechanisms that could form it. We measure a bipolar circumstellar structure with a position angle of 42.3$\pm 7.0^\circ$. We observe asymmetries in the form of hot spots in the north east of …

Towards Automated Evolution Of Imperative Deep Learning Programs, Tatiana Castro-Vélez

Dissertations, Theses, and Capstone Projects

Software engineering (SE) is increasingly intersecting with data-centric domains such as machine learning (ML) and deep learning (DL). Similar to bugs in traditional software systems, defects can emerge in ML and DL systems. ML, including DL, systems are now widespread and rely on dynamic models defined by input data. Developers face the challenge of building dependable systems while addressing the demand for scalable software.

Efficiency is essential to support responsiveness with respect to ever-growing datasets. Traditional DL frameworks achieve scalability through deferred execution, enabling symbolic, graph-based deep neural network (DNN) computation. While efficient, this approach is error-prone, cumbersome, and difficult …

Reinforcement Learning Based Resource Management In Edge Systems, Motahare Mounesan

Dissertations, Theses, and Capstone Projects

Modern end-user applications that are highly compute- and data-intensive, while being extremely latency- and accuracy-sensitive, are increasingly reliant on distributed computing. This paradigm spans a range of architectures, from cloud computing to in-device processing. Cloud computing, though scalable, often incurs high latency and cost, constraints that are particularly problematic for time-sensitive applications. In contrast, in-device computing on end or IoT devices is limited by resource constraints, making it inadequate for many complex workloads. Edge computing presents a compelling alternative by bringing computation closer to data sources, thereby reducing end-to-end latency and improving responsiveness. However, the inherent decentralized and dynamic nature …

Probabilistic Modeling, Learnability And Uncertainty Estimation For Interaction Prediction In Movie Rating Datasets, Jennifer Poernomo, Nicole Gabrielle Lee Tan, Rodrigo Alves, Antoine Ledent

Research Collection School Of Computing and Information Systems

In this paper, we examine the hypothesis that the interactions recorded in many Recommendation Systems datasets are distributed according to a low-rank distribution, i.e. a mixture of factorizable distributions. Surprisingly, we find that on several popular datasets, a simple non-negative matrix factorization method equals or outperforms more modern methods such as LightGCN, which indicates that the sampling distribution over interactions is indeed low-rank. Furthermore, we mathematically prove that low-rank distributions are learnable with a sparse number of observations (where m/n and r refer to the number of users/items and the non-negative rank respectively) both in terms of the total variation …

Managing Rumors On Electronic Interaction Platforms: How Management Responses Affect Investor Reaction, Runyu Wang, Zili Zhang, Keng Siau, Ziqiong Zhang

Research Collection School Of Computing and Information Systems

This study investigates how listed firms respond to investors’ rumor-related inquiries and examines the impact of these responses on investor reactions, as indicated by subsequent daily abnormal stock returns (ARs). Using a unique dataset of question-and-answer (Q&A) interactions from China’s major e-interaction platforms, established by the stock exchanges, our study provides insights into regulated firm-investor communications in a structured Q&A setting. Unlike informal social media channels, these platforms enable official responses from firm representatives, typically board secretaries, under direct regulatory oversight. By analyzing rumor-related Q&A pairs with regression models and several robustness checks, we find that firms can benefit from …

Boosting Symbolic Execution For Vulnerability Detection, Haoxin Tu

Dissertations and Theses Collection (Open Access)

Software systems written by humans tend to be unreliable and insecure, hence, bugs or vulnerabilities in them are inevitable. Symbolic execution has shown considerable potential in detecting diverse types of software bugs and also vulnerabilities that have severe security implications. However, existing symbolic execution engines still suffer from at least three fundamental limitations in memory modeling, path exploration, and structured input generation, which significantly impede existing engines from efficiently and effectively detecting software bugs and vulnerabilities.

The objective of this dissertation is to boost existing symbolic execution engines by designing a new memory model, two new path exploration strategies, and …

Memory-Efficient Graph Processing On Gpus: Reducing Intermediate Data Structure Overhead, Chang Ye

Dissertations and Theses Collection (Open Access)

The increasing scale of real-world graphs in domains such as fraud detection, community detection, and biological analysis demands high-throughput, memory-efficient graph processing solutions. GPUs offer massive parallelism for accelerating such workloads, and numerous frameworks have been developed to leverage their computational power. These frameworks primarily focus on optimizing scheduling to better align graph processing with GPU architectures. It performs well for algorithms with low memory demands, such as BFS, SSSP, and PageRank. However, for algorithms that require substantial memory, such as label propagation, and subgraph counting, the limited memory capacity of GPUs often becomes a significant bottleneck.

This dissertation addresses …

Recurrent Autoregressive Linear Model For Next-Basket Recommendation, Tereza Zmeskalova, Antoine Ledent, Martin Spisak, Pavel Kordik, Rodrigo Alves

Research Collection School Of Computing and Information Systems

Next-basket recommendation aims to predict the (sets of) items that a user is most likely to purchase during their next visit, capturing both short-term sequential patterns and long-term user preferences. However, effectively modeling these dynamics remains a challenge for traditional methods, which often struggle with interpretability and computational efficiency, particularly when dealing with intricate temporal dependencies and inter-item relationships. In this paper, we propose ReALM, a Recurrent Autoregressive Linear Model that explicitly captures temporal item-to-item dependencies across multiple time steps. By leveraging a recurrent loss function and a closed-form optimization solution, our approach offers both interpretability and scalability while maintaining …

An Efficient Security-Enhanced Accountable Access Control For Named Data Networking, Jianfei Sun, Yuxian Li, Xuehuan Yang, Guomin Yang, Robert H. Deng

Research Collection School Of Computing and Information Systems

Named Data Networking (NDN) is embraced as the crucial implementation of Information-Centric Networking (ICN), enhancing content distribution and caching efficiency through edge routers. However, existing NDN architectures face significant security and privacy challenges, including: (a) a lack of secure and efficient access control; (b) inadequate support for flexible and selective content management by content publishers; (c) insufficient implementation of accountability and privilege revocation mechanisms. To handle these challenges, we propose ESAS, the first-ever Efficient Security-enhanced Accountable Access Control Scheme for NDN. Specifically, our ESAS incorporates anonymous authentication using group signatures at network routers to prevent unauthorized access, employs key-aggregation-based access …

Lighttransfer: Your Long-Context Llm Is Secretly A Hybrid Model With Effortless Adaptation, Xuan Zhang, Fengzhuo Zhang, Cunxiao Du, Chao Du, Tianyu Pang, Wei Gao, Min Lin

Research Collection School Of Computing and Information Systems

Scaling language models to handle longer contexts introduces substantial memory challenges due to the growing cost of key-value (KV) caches. Motivated by the efficiency gains of hybrid models and the broad availability of pretrained large transformer backbones, we explore transitioning transformer models into hybrid architectures for a more efficient generation. In this work, we propose LightTransfer, a lightweight method that transforms models such as LLaMA into hybrid variants. Our approach identifies lazy layers -- those focusing on recent or initial tokens -- and replaces their full attention with streaming attention. This transformation can be performed without any training for long-context …

Ponzilens+: Visualizing Bytecode Actions For Smart Ponzi Scheme Identification, Xiaolin Wen, Tai D. Nguyen, Shaolun Ruan, Qiaomu Shen, Jun Sun, Feida Zhu, Yong Wang

Research Collection School Of Computing and Information Systems

With the prevalence of smart contracts, smart Ponzi schemes have become a common fraud on blockchain and have caused significant financial loss to cryptocurrency investors in the past few years. Despite the critical importance of detecting smart Ponzi schemes, a reliable and transparent identification approach adaptive to various smart Ponzi schemes is still missing. To fill the research gap, we first extract semantic-meaningful actions to represent the execution behaviors specified in smart contract bytecodes, which are derived from a literature review and in-depth interviews with domain experts. We then propose PonziLens+, a novel visual analytic approach that provides an intuitive …

A Difficult Act To Maintain, J.G. Allen

Research Collection Yong Pung How School Of Law

In "Where is Singapore's AI regulation headed?" (Issues, Summer 2025), Manoj Harjani starts by recounting its delicate balancing act in governing artificial intelligence. The city-state is able to maintain credibility on the global stage while remaining pragmatically grounded in technical experimentation, maintaining policy realism without overcommitting to rigid legal frameworks. The choice to adopt a regulatory “light touch” is often framed as a way to maintain flexibility, avoid overregulation, and enable innovation. But this obscures the extent to which this posture is itself a political and economic settlement—one that reinforces Singapore’s position in the global digital economy by facilitating capital …

Apidocbooster: An Extract-Then-Abstract Framework Leveraging Large Language Models For Augmenting Api Documentation, Chengran Yang, Jiakun Liu, Bowen Xu, Christoph Treude, Yunbo Lyu, Junda He, Ming Li, David Lo

Research Collection School Of Computing and Information Systems

API documentation is often the most trusted resource for programming. Many approaches have been proposed to augment API documentation by summarizing complementary information from external resources such as Stack Overflow. Existing extractive-based summarization approaches excel in producing faithful summaries that accurately represent the source content without input length restrictions. Nevertheless, they suffer from inherent readability limitations. On the other hand, our empirical study on the abstractive-based summarization method, i.e., GPT-4, reveals that GPT-4 can generate coherent and concise summaries but presents limitations in terms of informativeness and faithfulness. We introduce APIDocBooster, an extract-then-abstract framework that seamlessly fuses the advantages of …

Map As A By-Product: Collective Landmark Mapping From Imu Data And User-Provided Texts In Situated Tasks, Ryo Yonetani, Kotaro Hara

Research Collection School Of Computing and Information Systems

This paper presents Collective Landmark Mapper, a novel map-as-a-by-product system for generating semantic landmark maps of indoor environments. Consider users engaged in situated tasks that require them to navigate these environments and regularly take notes on their smartphones. Collective Landmark Mapper exploits the smartphone's IMU data and the user's free text input during these tasks to identify a set of landmarks encountered by the user. The identified landmarks are then aggregated across multiple users to generate a unified map representing the positions and semantic information of all landmarks. In developing the proposed system, we focused specifically on retail applications and …

Conv4rec: A 1‑By‑1 Convolutional Autoencoder For User Profiling Through Joint Analysis Of Implicit And Explicit Feedbacks, Antoine Ledent, Petr Kasalický, Rodrigo Alves, Hady Wirawan Lauw

Research Collection School Of Computing and Information Systems

We introduce a new convolutional autoencoder architecture for user modeling and recommendation tasks with several improvements over the state of the art. First, our model has the flexibility to learn a set of associations and combinations between different interaction types in a way that carries over to each user and item. Second, our model is able to learn jointly from both the explicit ratings and the implicit information in the sampling pattern (which we refer to as ”implicit feedback”). It can also make separate predictions for the probability of consuming content and the likelihood of granting it a high rating …

Finding Safety Violations Of Ai-Enabled Control Systems Through The Lens Of Synthesized Proxy Programs, Jieke Shi, Zhou Yang, Junda He, Bowen Xu, Dongsun Kim, Donggyun Han, David Lo

Research Collection School Of Computing and Information Systems

Given the increasing adoption of modern AI-enabled control systems, ensuring their safety and reliability has become a critical task in software testing. One prevalent approach to testing control systems is falsification, which aims to find an input signal that causes the control system to violate a formal safety specification using optimization algorithms. However, applying falsification to AI-enabled control systems poses two significant challenges: (1) it requires the system to execute numerous candidate test inputs, which can be time-consuming, particularly for systems with AI models that have many parameters, and (2) multiple safety requirements are typically defined as a conjunctive specification, …

Robface: A Test Suite For Efficient Robustness Evaluation Of Face Recognition Systems, Ruihan Zhang, Jun Sun

Research Collection School Of Computing and Information Systems

Face recognition is a widely used authentication technology in practice, where robustness is required. It is thus essential to have an efficient and easy-to-use method for evaluating the robustness of (possibly third-party) trained face recognition systems. Existing approaches to evaluating the robustness of face recognition systems are either based on empirical evaluation (e.g., measuring attacking success rate using state-of-the-art attacking methods) or formal analysis (e.g., measuring the Lipschitz constant). While the former demands significant user efforts and expertise, the latter is extremely time-consuming. In pursuit of a comprehensive, efficient, easy-to-use, and scalable estimation of the robustness of face recognition systems, …

Reimagining Academic Assessment In The Age Of Ai, Matthew Hammerton

Research Collection School of Social Sciences

In ‘Reimagining Academic Assessment in the Age of AI’, Matthew Hammerton examines the challenges and opportunities posed by generative AI for higher education assessment. He critiques common responses like banning AI, reverting to in-class exams, or abandoning essays altogether, arguing that they fail to preserve the deeper pedagogical goals of higher order, independent thinking. Instead, Hammerton proposes a guiding principle of intellectual responsibility: students should be accountable for explaining and defending each major choice in their work—regardless of whether they use AI tools. To operationalise this, he advocates for reintegrating oral examinations (‘vivas’) alongside written essays. In this model, students …