Sitemap
A list of all the posts and pages found on the site. For you robots out there, there is an XML version available for digesting as well.
Pages
Posts
Future Blog Post
Published:
This post will show up by default. To disable scheduling of future posts, edit config.yml and set future: false.
Blog Post number 4
Published:
This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.
Blog Post number 3
Published:
This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.
Blog Post number 2
Published:
This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.
Blog Post number 1
Published:
This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.
portfolio
Portfolio item number 1
Short description of portfolio item number 1
Portfolio item number 2
Short description of portfolio item number 2
publications
Deep Learning Models for Fault Detection and Diagnosis in Photovoltaic Modules Manufacture
Published in IEEE Conference on Artificial Intelligence (CAI 2023), 2023
Abstract
The usage of photovoltaic (PV) systems has experienced exponential growth. This growth, however, places gargantuan pressure on the solar energy industry’s manufacturing sector and subsequently begets issues associated with the quality of PV systems, especially the PV module. Currently, fault detection and diagnosis (FDD) are challenging due to many factors including but not limited to requirements of sophisticated measurement instruments and experts. Recent advances in deep learning (DL) have proven its feasibility in image classification and object detection. Thus, DL can be extended to visual fault detection using data generated by electroluminescence (EL) imaging instruments. Here, the authors propose an in-depth approach to exploratory data analysis of EL data and several techniques based on supervised learning to detect and diagnose visual faults and defects presented in a module.
Recommended citation: Nguyen-Vinh, K., Vo-Huynh, Q.-N., Nguyen-Minh, K., & Hoang, M. (2023). Deep Learning Models for Fault Detection and Diagnosis in Photovoltaic Modules Manufacture. IEEE CAI 2023.
Download Paper
Case Study: Utilising Deep Learning Models for Fault Detection and Diagnosis of Photovoltaic Modules
Published in Information Systems Research in Vietnam, Volume 2, Springer, 2023
Abstract
Renewable energy sources have long been considered to be the sole alternatives to fossil fuels. Consequently, the usage of PV systems has experienced exponential growth. This growth, however, places gargantuan pressure on the solar energy industry’s manufacturing sector and subsequently begets issues associated with the quality of PV systems, especially the PV module, which is the systems’ most crucial component. Currently, fault detection and diagnosis are challenging due to many factors including but not limited to requirements of sophisticated measurement instruments and experts. Recent advances in deep learning have proven its feasibility in image classification and object detection. Thus, deep learning can be extended to visual fault detection using data generated by electroluminescence imaging instruments. Here, the authors propose an in-depth approach to exploratory data analysis of electroluminescence data, as well as several techniques based on both supervised and unsupervised learning to detect and diagnose visual faults and defects presented in a module.
Recommended citation: Nguyen-Vinh, K., Vo-Huynh, Q.-N., Nguyen-Minh, K., & Hoang, M. (2023). Case Study: Utilising Deep Learning Models for Fault Detection and Diagnosis of Photovoltaic Modules. Springer, Information Systems Research in Vietnam, Volume 2.
Download Paper
Counterfactual Explanations and Algorithmic Recourses for Machine Learning: A Review
Published in ACM Computing Surveys, 2024
Abstract
Machine learning plays a role in many deployed decision systems, often in ways that are difficult or impossible to understand by human stakeholders. Explaining, in a human-understandable way, the relationship between the input and output of machine learning models is essential to the development of trustworthy machine learning based systems. A burgeoning body of research seeks to define the goals and methods of explainability in machine learning. In this article, we seek to review and categorize research on counterfactual explanations, a specific class of explanation that provides a link between what could have happened had input to a model been changed in a particular way. Modern approaches to counterfactual explainability in machine learning draw connections to the established legal doctrine in many countries, making them appealing to fielded systems in high-impact areas such as finance and healthcare. Thus, we design a rubric with desirable properties of counterfactual explanation algorithms and comprehensively evaluate all currently proposed algorithms against that rubric. Our rubric provides easy comparison and comprehension of the advantages and disadvantages of different approaches and serves as an introduction to major research themes in this field. We also identify gaps and discuss promising research directions in the space of counterfactual explainability.
Recommended citation: Verma, S., Boonsanong, V., Hoang, M., Hines, K. E., Dickerson, J. P., & Shah, C. (2024). Counterfactual Explanations and Algorithmic Recourses for Machine Learning: A Review. ACM Computing Surveys.
Download Paper
DenseTransXR: A Hybrid Transformer-Based Approach for Chest X-ray Abnormality and Emerging Disease Detection
Published in FPT International Conference on Emerging Trends in Computing (FETC 2025), 2025
Abstract
The COVID-19 pandemic has caused enormous challenges for medical experts in identifying the abnormalities from radiography images due to the lack of pre-labeled training data. Consequently, the demand for comprehensive labeled datasets and effective analytical methods for radiographic analysis increased dramatically, particularly in scenarios involving previously unseen diseases. In this study, we proposed DenseTransXR, a hybrid architecture that integrates the DenseNet-121 feature extractor with the global contextual modeling of Transformer encoders via Multi-Head Self Attention. Our model achieved an AUC of 0.812 for multi-label abnormality detection on NIH ChestX-ray14, outperforming both CNN-based and hybrid baselines, as well as an AUC of 0.755 and a recall of 0.919 for zero-shot COVID-19 detection on the COVIDx CXR-4 dataset, demonstrating strong generalization in identifying previously unseen diseases. These results showed the potentials of hybrid Transformer models in improving diagnostic performance and providing scalable solutions for future pandemics.
Recommended citation: Hoang, M. A., & Phan, H. L. M. (2025). DenseTransXR: A Hybrid Transformer-Based Approach for Chest X-ray Abnormality and Emerging Disease Detection. FETC 2025.
Download Paper
Forecasting Urban Energy for Net-Zero Smart Cities: A Hybrid Attention-based Approach
Published in FPT International Conference on Emerging Trends in Computing (FETC 2025), 2025
Abstract
The transition towards net-zero smart cities required complex and agile energy forecasting models to manage nonlinear energy consumption patterns and incorporate data from various sources in real-time. AI has emerged as a transformative tool in this domain, leveraging Deep Learning techniques to provide highly accurate and adaptive energy predictions. This study proposed two hybrid deep learning architectures combining CNN for short-term feature extraction, LSTM for long-range temporal modeling, with an Attention mechanism for dynamic time-step weighting to enhance predictive performance in both single-step and multi-step energy forecasting. Evaluated on the ENTSO-E dataset, the proposed CNN-LSTM-Att architectures significantly outperformed other baselines, with CNN-LSTM-Att.v2 achieving the best single-step forecasting results with RMSE of 106.96 MW, reducing MAPE to 1.07%, and CNN-LSTM-Att.v1 achieving the best multi-step day-ahead forecasting results with RMSE of 438.11 MW and reducing MAPE to 4.18%. These experimental results highlighted the importance of spatial-temporal feature extraction and attention-driven modeling in delivering robust forecasts for sustainable urban energy management.
Recommended citation: Hoang, M. A., Hoang, T. D. T., Phan, T. P., & Nguyen-Vinh, K. (2025). Forecasting Urban Energy for Net-Zero Smart Cities. FETC 2025.
Download Paper
AI-Based Energy Forecasting for Net Zero Smart Cities
Published in IEEE ICECIE 2025, 2025
Abstract
The transition toward net-zero smart cities necessitates advanced energy forecasting methodologies to enhance sustainability, optimize energy distribution, and balance supply-demand dynamics. Artificial Intelligence (AI) has emerged as a transformative tool in this domain, leveraging machine learning (ML) and deep learning (DL) techniques to provide highly accurate and adaptive energy predictions. This paper explores AI-driven energy forecasting frameworks, emphasizing the integration of real-time data from IoT-enabled smart grids, weather prediction models, and historical energy consumption patterns. Key AI methodologies, including time-series forecasting models, reinforcement learning strategies, and hybrid AI techniques, are examined for their efficacy in optimizing energy management. Additionally, this study addresses major challenges such as data integrity, model scalability, and cybersecurity risks associated with AI-driven energy systems. Through case studies and comparative analyses, we highlight the tangible benefits of AI in achieving net-zero energy objectives, offering insights into future advancements for sustainable urban development.
Recommended citation: Hoang, M., Nguyen-Vinh, K., & Vo, M.-T. (2025). AI-Based Energy Forecasting for Net Zero Smart Cities. IEEE ICECIE 2025.
Download Paper
AI-Driven Methods for Eliminating Harmonic Oscillations in Inverter-Connected Systems
Published in IEEE ICECIE 2025, 2025
Abstract
The increasing integration of inverter-based systems in modern power grids has introduced significant power quality challenges, particularly harmonic oscillations that can degrade system stability and efficiency. This paper proposes an AI-driven adaptive detection approach to identify and mitigate harmonic oscillations in inverter-connected systems. By leveraging artificial intelligence methods, the system can dynamically extract dominant harmonic frequencies and activate appropriate harmonic traps to suppress oscillations in real time. The proposed method explores various AI techniques, including wavelet transform-based neural networks, convolutional neural networks (CNNs), and recurrent neural networks (RNNs), to enhance accuracy and adaptability. Comparative evaluations with traditional Fourier and wavelet-based approaches demonstrate the effectiveness of the AI-driven solution in achieving faster and more precise harmonic suppression. This research contributes to the advancement of intelligent power quality management, offering a scalable and efficient approach to ensure stable inverter-based power systems.
Recommended citation: Nguyen-Vinh, K., Hoang, M., & Gono, R. (2025). AI-Driven Methods for Eliminating Harmonic Oscillations in Inverter-Connected Systems. IEEE ICECIE 2025.
Download Paper
EfficientSkinTrans: Enhancing Dermatological Analysis Through a Hybrid ViT Network
Published in IEEE CWCC 2026, 2026
Abstract
Accurate diagnosis of skin lesions remains challenging due to high visual variability and overlap among dermatological conditions. While prior work has explored combining CNNs with transformer encoders, many existing hybrids rely on heavy or redundant backbones that limit generalization. To address this, we proposed EfficientSkinTrans, a lightweight yet powerful hybrid architecture that integrates an EfficientNet-style convolutional encoder with a compact transformer module to jointly capture local texture patterns and long-range contextual dependencies. Experiments on the ISIC-2019 dataset showed that EfficientSkinTrans achieved strong performance across all evaluation metrics, outperforming conventional Transformer-based and recent hybrid approaches. The model also demonstrated notable zero-shot generalization to unseen diseases, including Monkeypox images from the MSLD v2.0 dataset, despite receiving no prior exposure to this condition. To ensure clinical trustworthiness, we also employed Grad-CAM++ to confirm that the model consistently attended to medically relevant lesion regions. These results highlighted EfficientSkinTrans as a reliable, interpretable, and generalizable solution for AI-assisted dermatological diagnosis, suitable both for common conditions and emerging infectious diseases.
Recommended citation: Hoang, M. A., Phan, H. L. M., Vo, N. N., & Nguyen-Vinh, K. (2026). EfficientSkinTrans: Enhancing Dermatological Analysis Through a Hybrid ViT Network. IEEE CWCC 2026.
Download Paper
talks
Deep Learning Models for Fault Detection and Diagnosis in Photovoltaic Modules Manufacture
Published:
In this talk, I highlighted the growing need for reliable fault detection and diagnosis (FDD) in photovoltaic (PV) manufacturing as global solar adoption accelerates. Traditional FDD relies on specialized instruments and expert analysis, creating bottlenecks in quality assurance. I introduced a deep learning–based approach that leverages electroluminescence (EL) imaging to automatically identify visual defects in PV modules. The presentation covered exploratory data analysis of EL images and several supervised learning techniques designed to detect and classify module faults, demonstrating how modern DL methods can significantly enhance PV quality control and support scalable, high-precision manufacturing.
Case Study: Utilizing Deep Learning Models for Fault Detection and Diagnosis of Photovoltaic Modules to Improve Solar Energy Constructions’ O&M Activities Quality
Published:
I delivered guest lectures at RMIT University Vietnam for Undergraduate Courses EEET2602 and EEET2623 on Deep Learning approaches for photovoltaic (PV) materials and semiconductor systems, covering model architectures, PV fault detection, and real-world applications in sustainable energy engineering. The lectures can be found here (2024) and here (2025).
DenseTransXR: A Hybrid Transformer-Based Approach for Chest X-ray Abnormality and Emerging Disease Detection
Published:
In this talk, I presented DenseTransXR, a hybrid architecture that combines DenseNet-121 with Transformer encoders to address the shortage of labeled radiography data during the COVID-19 pandemic. The model achieved an AUC of 0.812 on NIH ChestX-ray14 for multi-label abnormality detection and demonstrated strong zero-shot generalization on COVIDx CXR-4 with an AUC of 0.755 and a recall of 0.919 for COVID-19 detection. These results highlight the capability of hybrid Transformer models to improve radiographic diagnosis and handle previously unseen diseases effectively.
Forecasting Urban Energy for Net-Zero Smart Cities: A Hybrid Attention-based Approach
Published:
In this talk, I presented two hybrid deep learning architectures for net-zero smart city energy forecasting, integrating CNNs for short-term feature extraction, LSTMs for long-range temporal modeling, and an Attention mechanism for dynamic weighting of time-step importance. Evaluated on the ENTSO-E dataset, the proposed models outperformed all baselines, with CNN-LSTM-Att.v2 achieving the best single-step results (RMSE 106.96 MW, MAPE 1.07%) and CNN-LSTM-Att.v1 delivering the strongest multi-step day-ahead performance (RMSE 438.11 MW, MAPE 4.18%). These findings demonstrate the effectiveness of attention-driven spatiotemporal modeling for accurate and scalable energy forecasting in sustainable smart cities. The talk can be found here.
AI-Driven Adaptive Detection for Eliminating Harmonic Oscillations in Inverter-Connected Systems
Published:
In this talk, I presented an AI-driven adaptive approach for detecting and mitigating harmonic oscillations in inverter-connected power systems, an increasingly critical issue as modern grids integrate more inverter-based resources. The method leverages AI techniques, including wavelet-based neural networks, CNNs, and RNNs, to dynamically extract dominant harmonic frequencies and trigger real-time harmonic traps for suppression. Compared with traditional Fourier and wavelet-based methods, the AI-driven models achieved faster and more accurate harmonic detection, demonstrating a scalable and effective solution for enhancing power quality and ensuring stable inverter-based power system operation. The talk can be found here.
AI-based Energy Forecasting for Net Zero Smart Cities
Published:
In this talk, I discussed how AI-driven energy forecasting is accelerating the transition toward net-zero smart cities by enabling more accurate, adaptive, and efficient energy management. The presentation explored key AI methodologies, including advanced time-series models, deep learning strategies, and hybrid AI techniques, designed to integrate real-time IoT grid data, weather information, and historical consumption patterns. I highlighted major challenges such as data integrity, model scalability, and cybersecurity, while demonstrating through case studies how AI significantly improves energy planning, distribution optimization, and supply–demand balancing. The talk emphasized AI’s growing role as a foundational technology for achieving sustainable, net-zero urban development. The talk can be found here.
EfficientSkinTrans: Enhancing Dermatological Classification and Analysis Through a Hybrid ViT Network
Published:
In this talk, I introduced EfficientSkinTrans, a lightweight hybrid model designed to improve skin lesion diagnosis by combining an EfficientNet-style convolutional encoder with a compact transformer module to capture both fine-grained textures and long-range contextual cues. Evaluated on ISIC-2019, the model outperformed conventional transformer-based and hybrid approaches, while also demonstrating strong zero-shot generalization to unseen diseases such as Monkeypox from the MSLD v2.0 dataset. To support clinical trust, Grad-CAM++ visualizations confirmed that the model consistently focused on medically relevant lesion regions. Overall, EfficientSkinTrans provides a reliable, interpretable, and generalizable framework for AI-assisted dermatological diagnosis.
teaching
Swinburne University of Technology
Full-time Lecturer, Ho Chi Minh City, 2024
- COS10022: Introduction to Data Science (Fall 2024, Summer 2025)
- COS20007: Object Oriented Programming (Fall 2024, Spring 2025)
- COS20028: Big Data Architecture and Application (Summer 2025, Fall 2025)
- COS30045: Data Visualization (Fall 2025, Spring 2026)
- COS30082: Applied Machine Learning (Fall 2024, Summer 2025, Fall 2025)
- COS40007: Artificial Intelligence for Engineering (Spring 2026)
- STA10003: Foundations of Statistics (Spring 2026)
FPT University
Part-time Lecturer, Ho Chi Minh City, 2025
- DBM302m: Data Mining (Summer 2025)
- DAT301m: AI Development with TensorFlow (Summer 2025)
Asia University
Part-time Lecturer, Ho Chi Minh City, 2025
- AU005: Introduction to Information Technology (Fall 2025)