Dr. Chao Yuan | Machine Learning | Best Researcher Award

Associate Professor at Guangzhou University, China

Dr. Chao Yuan is a postdoctoral researcher at the School of Mathematics and Information Science, Guangzhou University, and a visiting scholar at Durham University, UK. He earned his Ph.D. in Computer Science and Technology from China Agricultural University in 2022. His research focuses on machine learning, particularly robust metric learning and nonlinear classification methods. Dr. Yuan has authored over fifteen high-impact journal articles in top-tier journals such as Knowledge-Based Systems, Neural Networks, and Information Sciences. His contributions span both theoretical advancements and practical implementations in areas like image denoising, signal reconstruction, and pattern classification. Known for his strong analytical mindset, innovative thinking, and team collaboration, Dr. Yuan is recognized for delivering results in complex research environments. With a clear vision for interdisciplinary exploration, he aims to bridge cutting-edge learning models with real-world intelligent systems. His career reflects dedication to academic excellence, continuous learning, and impactful scientific discovery.

Professional Profile:

Scopus

Education Background

Dr. Yuan earned his B.Sc. in Information and Computing Science from Weinan Normal University (2014), his M.Sc. in Computational Mathematics from Xi’an Polytechnic University (2018), and his Ph.D. in Computer Science and Technology from China Agricultural University (2022). His academic training bridges mathematics, computing, and artificial intelligence. During his Ph.D., he specialized in machine learning algorithms, robust metric learning, and classification techniques. His education laid a strong theoretical and computational foundation, equipping him with skills in optimization, signal analysis, and modeling. He has actively participated in research during all academic phases, contributing to publications and national projects. His academic journey reflects continuous growth from applied mathematics to cutting-edge intelligent computing.

Professional Development

Dr. Yuan is currently a postdoctoral researcher at Guangzhou University and a visiting scholar at Durham University (2023–2024) under the Guangdong Young Talents Program. He previously participated in multiple national projects during his doctoral research, focusing on sparse coding, manifold learning, and image set classification. He has experience in algorithm development, scientific publishing, and interdisciplinary collaboration. His professional work spans robust AI models, lightweight architectures for IoT, and biologically inspired computation. At Durham, he is currently researching swarm intelligence and robotic systems. Dr. Yuan brings practical innovation and academic rigor to his work, with a commitment to applied research and impactful discoveries.

Research Focus

Dr. Yuan’s research interests include machine learning, robust classification, nonlinear metric learning, sparse representation, image denoising, and manifold learning. He focuses on correntropy-based techniques and adaptive learning methods for noise-tolerant AI. His work also explores Riemannian manifold approaches, lightweight deep networks, and swarm intelligence for autonomous systems. He is passionate about developing efficient and interpretable models for real-world tasks, especially in constrained environments like IoT. Dr. Yuan is currently researching intelligent swarm systems, combining bio-inspired algorithms with AI. His long-term goal is to bridge theory and application, creating robust, scalable, and generalizable intelligent systems.

Author Metrics:

Awards and Honors:

1.  Mixture correntropy-based robust distance metric learning for classification

~ Authors: Chao Yuan, Changsheng Zhou, Jigen Peng, Haiyang Li
~ Journal: Knowledge-Based Systems, 2024, Volume 295, Article 111791, Pages 1–20
~Impact Factor: 8.8 (CAS Zone 1)

Summary:
This paper proposes a novel distance metric learning algorithm using mixture correntropy to handle non-Gaussian noise and outliers in classification tasks. It demonstrates improved robustness and accuracy compared to existing methods, especially in noisy and real-world datasets.

2. Correntropy-based metric for robust twin support vector machine

~ Authors: Chao Yuan, Liming Yang, Ping Sun
~Journal: Information Sciences, 2021, Volume 545(1), Pages 82–101
~ Impact Factor: 8.1 (CAS Zone 1)

Summary:
This work integrates correntropy into Twin Support Vector Machines (TWSVM), resulting in a classifier that is more resistant to noise and outliers. The model exhibits better generalization and classification performance on challenging datasets.

3. Robust twin extreme learning machines with correntropy-based metric

~ Authors: Chao Yuan, Liming Yang
~ Journal: Knowledge-Based Systems, 2021, Volume 214, Article 106707, Pages 1–15
~Impact Factor: 8.8 (CAS Zone 1)

Summary:
The authors enhance Twin Extreme Learning Machines (TELM) by incorporating a correntropy-based loss function, making them more robust for classification tasks in the presence of noisy labels and outliers.

4. Capped L2,p-norm metric based robust least squares twin support vector machine for pattern classification

~  Authors: Chao Yuan, Liming Yang
~ Journal: Neural Networks, 2021, Volume 142, Pages 457–478
~ Impact Factor: 7.8 (CAS Zone 1)

Summary:
This paper introduces a capped L2,p-norm-based metric into the Least Squares Twin SVM framework, enhancing robustness by mitigating the influence of noisy and redundant samples. It shows superior classification accuracy across benchmark datasets.

5. Large margin projection-based multi-metric learning for classification

~  Authors: Chao Yuan, Liming Yang
~  Journal: Knowledge-Based Systems, 2022, Volume 243, Article 108481, Pages 1–15

Summary:  This research presents a multi-metric learning approach based on large-margin projections that dynamically adjusts distance metrics for different data subspaces. The method significantly enhances classification accuracy and adaptability to diverse data distributions.

Conclusion