Kimleang Kea
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Kimleang Kea has a total of 15 co-authors. Here are the top co-authors by number of shared publications:
- Youngsun Han
Pukyong National University
8 shared publications - Young-Jae Min
Department of Electric and Electronic Engineering, Halla University, Wonju, South Korea
3 shared publications Chansreynich Huot
Department of AI Convergence, Pukyong National University, Nam-gu, Busan, South Korea
2 shared publications- Tae-Kyung Kim
Department of Management Information Systems, Chungbuk National University, Seowon-gu, Cheongju-si, Chungcheongbuk-do, South Korea
2 shared publications - Chansreynich Huot
Department of AI Convergence, Pukyong National University, Nam-gu, Busan, South Korea
1 shared publication - Dongmin Kim
Pukyong National University
1 shared publication
Latest Publications
- Open Access
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 113766
Corrections to “A Federated Learning Approach for Efficient Anomaly Detection in Electric Power Steering Systems”
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 113766
Open AccessCorrections to “A Federated Learning Approach for Efficient Anomaly Detection in Electric Power Steering Systems”
- Open Access
- 30 October 2025
- Published by IOP Publishing in Machine Learning: Science and Technology
- Vol. 6 (4) , 045027
Enhancing a convolutional autoencoder with a quantum approximate optimization algorithm for image noise reduction
Image denoising is essential for removing noise in images caused by electric device malfunctions or other factors during image acquisition. It ensures the preservation of image quality and accurate interpretation. Many convolutional autoencoder algorithms have proven effective in image denoising. Owing to their promising efficiency, quantum computers have gained popularity. This paper proposes a method, the quantum convolutional autoencoder (QCAE), which enhances traditional convolutional autoencoders by replacing their latent space with a quantum counterpart implemented via a quantum approximate optimization algorithm (QAOA)-inspired ansatz circuit. To enhance efficiency, we leveraged the advantages of the QAOA-incorporated parameter-shift rule to identify an optimized cost function, facilitating effective learning from data and gradient computation on an actual quantum computer. The proposed QCAE method outperformed its classical counterpart as it exhibited lower training loss and a higher structural similarity index (SSIM) value. QCAE also outperformed its classical counterpart in denoising the MNIST dataset by up to 40% in terms of SSIM value, confirming its enhanced capabilities in real-world applications. Evaluation of QAOA performance across different circuit configurations and layer variations showed that our technique outperformed other circuit designs by 25% on average.
- 30 October 2025
- Published by IOP Publishing in Machine Learning: Science and Technology
- Vol. 6 (4) , 045027
Open AccessEnhancing a convolutional autoencoder with a quantum approximate optimization algorithm for image noise reduction
Image denoising is essential for removing noise in images caused by electric device malfunctions or other factors during image acquisition. It ensures the preservation of image quality and accurate interpretation. Many convolutional autoencoder algorithms have proven effective in image denoising. Owing to their promising efficiency, quantum computers have gained popularity. This paper proposes a method, the quantum convolutional autoencoder (QCAE), which enhances traditional convolutional autoencoders by replacing their latent space with a quantum counterpart implemented via a quantum approximate optimization algorithm (QAOA)-inspired ansatz circuit. To enhance efficiency, we leveraged the advantages of the QAOA-incorporated parameter-shift rule to identify an optimized cost function, facilitating effective learning from data and gradient computation on an actual quantum computer. The proposed QCAE method outperformed its classical counterpart as it exhibited lower training loss and a higher structural similarity index (SSIM) value. QCAE also outperformed its classical counterpart in denoising the MNIST dataset by up to 40% in terms of SSIM value, confirming its enhanced capabilities in real-world applications. Evaluation of QAOA performance across different circuit configurations and layer variations showed that our technique outperformed other circuit designs by 25% on average.
- Open Access
- 10 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 98322
Corrections to “Leveraging Knapsack QAOA Approach for Optimal Electric Vehicle Charging”
Presents corrections to the paper, Corrections to “Leveraging Knapsack QAOA Approach for Optimal Electric Vehicle Charging”.
- 10 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 98322
Open AccessCorrections to “Leveraging Knapsack QAOA Approach for Optimal Electric Vehicle Charging”
Presents corrections to the paper, Corrections to “Leveraging Knapsack QAOA Approach for Optimal Electric Vehicle Charging”.
- Open Access
A Hybrid Quantum-Classical Model for Stock Price Prediction Using Quantum-Enhanced Long Short-Term Memory
The stock markets have become a popular topic within machine learning (ML) communities, with one particular application being stock price prediction. However, accurately predicting the stock market is a challenging task due to the various factors within financial markets. With the introduction of ML, prediction techniques have become more efficient but computationally demanding for classical computers. Given the rise of quantum computing (QC), which holds great promise for being exponentially faster than current classical computers, it is natural to explore ML within the QC domain. In this study, we leverage a hybrid quantum-classical ML approach to predict a company’s stock price. We integrate classical long short-term memory (LSTM) with QC, resulting in a new variant called QLSTM. We initially validate the proposed QLSTM model by leveraging an IBM quantum simulator running on a classical computer, after which we conduct predictions using an IBM real quantum computer. Thereafter, we evaluate the performance of our model using the root mean square error (RMSE) and prediction accuracy. Additionally, we perform a comparative analysis, evaluating the prediction performance of the QLSTM model against several other classical models. Further, we explore the impacts of hyperparameters on the QLSTM model to determine the best configuration. Our experimental results demonstrate that while the classical LSTM model achieved an RMSE of 0.0693 and a prediction accuracy of 0.8815, the QLSTM model exhibited superior performance, achieving values of 0.0602 and 0.9736, respectively. Furthermore, the QLSTM outperformed other classical models in both metrics.
Open AccessA Hybrid Quantum-Classical Model for Stock Price Prediction Using Quantum-Enhanced Long Short-Term Memory
The stock markets have become a popular topic within machine learning (ML) communities, with one particular application being stock price prediction. However, accurately predicting the stock market is a challenging task due to the various factors within financial markets. With the introduction of ML, prediction techniques have become more efficient but computationally demanding for classical computers. Given the rise of quantum computing (QC), which holds great promise for being exponentially faster than current classical computers, it is natural to explore ML within the QC domain. In this study, we leverage a hybrid quantum-classical ML approach to predict a company’s stock price. We integrate classical long short-term memory (LSTM) with QC, resulting in a new variant called QLSTM. We initially validate the proposed QLSTM model by leveraging an IBM quantum simulator running on a classical computer, after which we conduct predictions using an IBM real quantum computer. Thereafter, we evaluate the performance of our model using the root mean square error (RMSE) and prediction accuracy. Additionally, we perform a comparative analysis, evaluating the prediction performance of the QLSTM model against several other classical models. Further, we explore the impacts of hyperparameters on the QLSTM model to determine the best configuration. Our experimental results demonstrate that while the classical LSTM model achieved an RMSE of 0.0693 and a prediction accuracy of 0.8815, the QLSTM model exhibited superior performance, achieving values of 0.0602 and 0.9736, respectively. Furthermore, the QLSTM outperformed other classical models in both metrics.
- Open Access
- 6 May 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 67525-67536
A Federated Learning Approach for Efficient Anomaly Detection in Electric Power Steering Systems
Electric power steering (EPS) has emerged as a valuable driver-assistance system. In an EPS system, an extensive amount of data collected from various sensors is analyzed to enhance the driving experience. Anomaly detection techniques have shown potential in ensuring the integrity of data patterns and detecting abnormalities to prevent adverse driving incidents, thus improving vehicle safety. However, traditional centralized anomaly detection methods require the collection of data from all EPS sensors, resulting in high communication network overhead. Herein, we propose an approach for anomaly detection using EPS data within a federated learning (FL) framework. Our approach exploits deep learning (DL) on time series data to achieve highly effective anomaly detection. This synergy of FL and DL enables all sensors in an EPS system to collaboratively train the anomaly detection model simultaneously. Our results demonstrate the feasibility of combining FL with DL anomaly detection in EPS systems, thus overcoming the limitations of the traditional centralized approach. The combined FL and DL model performs remarkably well, achieving an F1-Score of 99.01%, outperforming the 98.31%~98.89% achieved by the centralized approach. It also exhibits high generalizability by incorporating insights from all sensors to comprehensively understand diverse driving scenarios. This results in a significant reduction in error rates, ranging from 20% to 33.25%, compared to centralized methods. Additionally, the proposed model exhibits significant advantages over traditional methods, including reduced training time and communication overhead, while maintaining comparable anomaly detection accuracy performance.
- 6 May 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 67525-67536
Open AccessA Federated Learning Approach for Efficient Anomaly Detection in Electric Power Steering Systems
Electric power steering (EPS) has emerged as a valuable driver-assistance system. In an EPS system, an extensive amount of data collected from various sensors is analyzed to enhance the driving experience. Anomaly detection techniques have shown potential in ensuring the integrity of data patterns and detecting abnormalities to prevent adverse driving incidents, thus improving vehicle safety. However, traditional centralized anomaly detection methods require the collection of data from all EPS sensors, resulting in high communication network overhead. Herein, we propose an approach for anomaly detection using EPS data within a federated learning (FL) framework. Our approach exploits deep learning (DL) on time series data to achieve highly effective anomaly detection. This synergy of FL and DL enables all sensors in an EPS system to collaboratively train the anomaly detection model simultaneously. Our results demonstrate the feasibility of combining FL with DL anomaly detection in EPS systems, thus overcoming the limitations of the traditional centralized approach. The combined FL and DL model performs remarkably well, achieving an F1-Score of 99.01%, outperforming the 98.31%~98.89% achieved by the centralized approach. It also exhibits high generalizability by incorporating insights from all sensors to comprehensively understand diverse driving scenarios. This results in a significant reduction in error rates, ranging from 20% to 33.25%, compared to centralized methods. Additionally, the proposed model exhibits significant advantages over traditional methods, including reduced training time and communication overhead, while maintaining comparable anomaly detection accuracy performance.
- 23 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-3
Accelerating YOLO-based Real-time Object Detection via Foveated Rendering
Foveated rendering emulates human vision by prioritizing sharpness at the center while introducing peripheral blur outside this central area. We explore its potential for enhancing object detection performance by integrating it into various object detection methods. Our investigation demonstrates that foveated rendering effectively reduces detection time and false alarm detections, thus showcasing its utility in improving object detection performance. Experimental results indicate a reduction in detection time of up to 20% compared with object detection using conventional rendering techniques.
- 23 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-3
Accelerating YOLO-based Real-time Object Detection via Foveated Rendering
Foveated rendering emulates human vision by prioritizing sharpness at the center while introducing peripheral blur outside this central area. We explore its potential for enhancing object detection performance by integrating it into various object detection methods. Our investigation demonstrates that foveated rendering effectively reduces detection time and false alarm detections, thus showcasing its utility in improving object detection performance. Experimental results indicate a reduction in detection time of up to 20% compared with object detection using conventional rendering techniques.
- Open Access
- 29 September 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 11, 109964-109973
Leveraging Knapsack QAOA Approach for Optimal Electric Vehicle Charging
The electric vehicle (EV) industry is currently afflicted with inefficient charging systems. Considering the growing adoption of EVs, optimization strategies for efficient charging, and overcoming constraints such as limited power supply and extended waiting times, are required. The knapsack algorithm, a classical technique that maximizes value and capacity, enables the efficient utilization of the limited available power supply while minimizing waiting times in EV charging scenarios. However, the knapsack problem is notoriously NP-hard, making it difficult to find efficient solutions classically. In this paper, we propose an approach that leverages the quantum approximation optimization algorithm (QAOA) to resolve the EV charging problem using a knapsack-based formulation. By incorporating a knapsack problem constraint into the QAOA, we overcome the limitations of the original QAOA method and provide a potential solution to the knapsack problem. We extensively evaluate and analyze the effectiveness of our approach in finding optimal EV charging solutions in both noise-free simulations and noisy real quantum devices. The proposed approach achieves impressive approximation ratios of up to 100% and 50% in noise-free and noisy environments, respectively. Even with small circuit sizes, we confirm that our approach can find optimal solutions effectively.
- 29 September 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 11, 109964-109973
Open AccessLeveraging Knapsack QAOA Approach for Optimal Electric Vehicle Charging
The electric vehicle (EV) industry is currently afflicted with inefficient charging systems. Considering the growing adoption of EVs, optimization strategies for efficient charging, and overcoming constraints such as limited power supply and extended waiting times, are required. The knapsack algorithm, a classical technique that maximizes value and capacity, enables the efficient utilization of the limited available power supply while minimizing waiting times in EV charging scenarios. However, the knapsack problem is notoriously NP-hard, making it difficult to find efficient solutions classically. In this paper, we propose an approach that leverages the quantum approximation optimization algorithm (QAOA) to resolve the EV charging problem using a knapsack-based formulation. By incorporating a knapsack problem constraint into the QAOA, we overcome the limitations of the original QAOA method and provide a potential solution to the knapsack problem. We extensively evaluate and analyze the effectiveness of our approach in finding optimal EV charging solutions in both noise-free simulations and noisy real quantum devices. The proposed approach achieves impressive approximation ratios of up to 100% and 50% in noise-free and noisy environments, respectively. Even with small circuit sizes, we confirm that our approach can find optimal solutions effectively.
- Open Access
- 18 August 2023
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 18 (8) , e0290337
Enhancing anomaly detection in distributed power systems using autoencoder-based federated learning
The growing use of Internet-of-Things devices in electric power systems has resulted in increased complexity and flexibility, making monitoring power usage critical for effective system maintenance and detecting abnormal behavior. However, traditional anomalous power consumption detection methods struggle to handle the vast amounts of data generated by these devices. While deep learning and machine learning are effective in anomaly detection, they require significant amounts of training data collected on centralized servers. This centralized approach results in high response time delays and data leakage problems. To address these challenges, we propose an Autoencoder-based Federated Learning method that combines the AutoEncoder and Federated Learning networks to develop a high-accuracy algorithm for identifying anomalies of power consumption data in distributed power systems. The proposed method allows for decentralized training of anomaly detection models among IoT devices, reducing response time and eventually solving data leakage issues. Our experimental results demonstrate the effectiveness of the FLAE method in detecting anomalies without needing data transferring.
- 18 August 2023
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 18 (8) , e0290337
Open AccessEnhancing anomaly detection in distributed power systems using autoencoder-based federated learning
The growing use of Internet-of-Things devices in electric power systems has resulted in increased complexity and flexibility, making monitoring power usage critical for effective system maintenance and detecting abnormal behavior. However, traditional anomalous power consumption detection methods struggle to handle the vast amounts of data generated by these devices. While deep learning and machine learning are effective in anomaly detection, they require significant amounts of training data collected on centralized servers. This centralized approach results in high response time delays and data leakage problems. To address these challenges, we propose an Autoencoder-based Federated Learning method that combines the AutoEncoder and Federated Learning networks to develop a high-accuracy algorithm for identifying anomalies of power consumption data in distributed power systems. The proposed method allows for decentralized training of anomaly detection models among IoT devices, reducing response time and eventually solving data leakage issues. Our experimental results demonstrate the effectiveness of the FLAE method in detecting anomalies without needing data transferring.
- Open Access
A Deep Learning Approach to Detect Anomalies in an Electric Power Steering System
As anomaly detection for electrical power steering (EPS) systems has been centralized using model- and knowledge-based approaches, EPS system have become complex and more sophisticated, thereby requiring enhanced reliability and safety. Since most current detection methods rely on prior knowledge, it is difficult to identify new or previously unknown anomalies. In this paper, we propose a deep learning approach that consists of a two-stage process using an autoencoder and long short-term memory (LSTM) to detect anomalies in EPS sensor data. First, we train our model on EPS data by employing an autoencoder to extract features and compress them into a latent representation. The compressed features are fed into the LSTM network to capture any correlated dependencies between features, which are then reconstructed as output. An anomaly score is used to detect anomalies based on the reconstruction loss of the output. The effectiveness of our proposed approach is demonstrated by collecting sample data from an experiment using an EPS test jig. The comparison results indicate that our proposed model performs better in detecting anomalies, with an accuracy of 0.99 and a higher area under the receiver operating characteristic curve than other methods providing a valuable tool for anomaly detection in EPS.
Open AccessA Deep Learning Approach to Detect Anomalies in an Electric Power Steering System
As anomaly detection for electrical power steering (EPS) systems has been centralized using model- and knowledge-based approaches, EPS system have become complex and more sophisticated, thereby requiring enhanced reliability and safety. Since most current detection methods rely on prior knowledge, it is difficult to identify new or previously unknown anomalies. In this paper, we propose a deep learning approach that consists of a two-stage process using an autoencoder and long short-term memory (LSTM) to detect anomalies in EPS sensor data. First, we train our model on EPS data by employing an autoencoder to extract features and compress them into a latent representation. The compressed features are fed into the LSTM network to capture any correlated dependencies between features, which are then reconstructed as output. An anomaly score is used to detect anomalies based on the reconstruction loss of the output. The effectiveness of our proposed approach is demonstrated by collecting sample data from an experiment using an EPS test jig. The comparison results indicate that our proposed model performs better in detecting anomalies, with an accuracy of 0.99 and a higher area under the receiver operating characteristic curve than other methods providing a valuable tool for anomaly detection in EPS.