Haoyuan Li
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Haoyuan Li has a total of 9 co-authors. Here are the top co-authors by number of shared publications:
- Aaqib Saeed
Eindhoven University of Technology, Eindhoven, The Netherlands
1 shared publication C F Jia
Department of Endocrinology, Sir Run Run Shaw Hospital, Zhejiang University School of Medicine, Hangzhou 310016, China
1 shared publicationF P Zheng
Department of Endocrinology, Sir Run Run Shaw Hospital, Zhejiang University School of Medicine, Hangzhou 310016, China
1 shared publication- Jindong Wang
William & Mary, Williamsburg, USA
1 shared publication - Mathias Funk
Industrial Design, Eindhoven University of Technology, Eindhoven, Netherlands
1 shared publication - Qianqian Pan
College of Agriculture and Biotechnology, Zhejiang University, Hangzhou, China
1 shared publication
Latest Publications
- 21 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 12 (15) , 31245-31255
FAST: Federated Active Learning With Foundation Models for Communication-Efficient Sampling and Training
Federated Active Learning (FAL) has emerged as a promising framework to leverage large quantities of unlabeled data across distributed clients while preserving data privacy. However, real-world deployments remain limited by high annotation costs and communication-intensive sampling processes, particularly in a cross-silo setting, when clients possess substantial local datasets. This article addresses the crucial question: What is the best practice to reduce communication costs in human-in-the-loop learning with minimal annotator effort? Existing FAL methods typically rely on iterative annotation processes that separate active sampling from federated updates, leading to multiple rounds of expensive communication and annotation. In response, we introduce FAST, a two-pass FAL framework that harnesses foundation models for weak labeling in a preliminary pass, followed by a refinement pass focused exclusively on the most uncertain samples. By leveraging representation knowledge from foundation models and integrating refinement steps into a streamlined workflow, FAST substantially reduces the overhead incurred by iterative active sampling. Extensive experiments on diverse medical and natural image benchmarks demonstrate that FAST outperforms existing FAL methods by an average of 4.36% while reducing communication rounds eightfold under a limited 5% labeling budget.
- 21 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 12 (15) , 31245-31255
FAST: Federated Active Learning With Foundation Models for Communication-Efficient Sampling and Training
Federated Active Learning (FAL) has emerged as a promising framework to leverage large quantities of unlabeled data across distributed clients while preserving data privacy. However, real-world deployments remain limited by high annotation costs and communication-intensive sampling processes, particularly in a cross-silo setting, when clients possess substantial local datasets. This article addresses the crucial question: What is the best practice to reduce communication costs in human-in-the-loop learning with minimal annotator effort? Existing FAL methods typically rely on iterative annotation processes that separate active sampling from federated updates, leading to multiple rounds of expensive communication and annotation. In response, we introduce FAST, a two-pass FAL framework that harnesses foundation models for weak labeling in a preliminary pass, followed by a refinement pass focused exclusively on the most uncertain samples. By leveraging representation knowledge from foundation models and integrating refinement steps into a streamlined workflow, FAST substantially reduces the overhead incurred by iterative active sampling. Extensive experiments on diverse medical and natural image benchmarks demonstrate that FAST outperforms existing FAL methods by an average of 4.36% while reducing communication rounds eightfold under a limited 5% labeling budget.
- 1 February 2023
- Published by Wolters Kluwer Health in Zhonghua Nei Ke Za Zhi
- Vol. 62 (2) , 169-175
[Nomogram analysis of risk factors for diabetic neuropathy in primary type 2 diabetic mellitus patients in Zhejiang Province].
Objective: To investigate the risk factors of diabetic nephropathy (DN) in primary type 2 diabetes mellitus (T2DM) patients and to quantitatively analyze the risk of DN by nomogram modeling. Methods: A total of 1 588 primary T2DM patients from 17 townships and streets in Zhejiang Province were enrolled from June 2018 to August 2018 in this cross-sectional study, with an average age of (56.8±10.1) years (50.06% male) and a mean disease duration of 9 years. The clinical data, biochemical test results, and fundus photographs of all T2DM patients were collected, and logistic regression analysis was used to screen the risk factors of DN. Then, a nomogram model was used to quantitatively analyze the risk of DN. Results: DN occurred in 27.71% (440/1 588 cases) primary type 2 diabetes patients. Hemoglobin A1c (HbA1c) (OR=1.159, 95%CI 1.039-1.292), systolic blood pressure (OR=1.041, 95%CI 1.031-1.051), serum creatinine (Scr) (OR=1.011, 95%CI 1.004-1.017), serum globulin (GLOB) (OR=1.072, 95%CI 1.039-1.105), diabetic retinopathy (DR) (OR=1.463, 95%CI 1.073-1.996), education level of more than junior high school (OR=2.018, 95%CI 1.466-2.777), and moderate-intensity exercise (OR=0.751, 95%CI 0.586-0.961) were influencing factors of DN. Nomogram model analysis showed that the total score of each factor of DN ranged from 64-138 points, and the corresponding risk rate ranged from 0.1-0.9. The nomogram model also predicted a C-index value of 0.753 (95%CI 0.726-0.781) and an area under the receiver operating characteristic curve of DN of 0.753. Internal verification of the C-index reached 0.738. The model displayed medium predictive power and could be applied in clinical practice. Conclusions: HbA1c, systolic blood pressure, Scr, GLOB, DR, and more than a junior high school education are independent risk factors of DN. Nomogram modeling can more intuitively evaluate the risk of DN in primary T2DM patients.
- 1 February 2023
- Published by Wolters Kluwer Health in Zhonghua Nei Ke Za Zhi
- Vol. 62 (2) , 169-175
[Nomogram analysis of risk factors for diabetic neuropathy in primary type 2 diabetic mellitus patients in Zhejiang Province].
Objective: To investigate the risk factors of diabetic nephropathy (DN) in primary type 2 diabetes mellitus (T2DM) patients and to quantitatively analyze the risk of DN by nomogram modeling. Methods: A total of 1 588 primary T2DM patients from 17 townships and streets in Zhejiang Province were enrolled from June 2018 to August 2018 in this cross-sectional study, with an average age of (56.8±10.1) years (50.06% male) and a mean disease duration of 9 years. The clinical data, biochemical test results, and fundus photographs of all T2DM patients were collected, and logistic regression analysis was used to screen the risk factors of DN. Then, a nomogram model was used to quantitatively analyze the risk of DN. Results: DN occurred in 27.71% (440/1 588 cases) primary type 2 diabetes patients. Hemoglobin A1c (HbA1c) (OR=1.159, 95%CI 1.039-1.292), systolic blood pressure (OR=1.041, 95%CI 1.031-1.051), serum creatinine (Scr) (OR=1.011, 95%CI 1.004-1.017), serum globulin (GLOB) (OR=1.072, 95%CI 1.039-1.105), diabetic retinopathy (DR) (OR=1.463, 95%CI 1.073-1.996), education level of more than junior high school (OR=2.018, 95%CI 1.466-2.777), and moderate-intensity exercise (OR=0.751, 95%CI 0.586-0.961) were influencing factors of DN. Nomogram model analysis showed that the total score of each factor of DN ranged from 64-138 points, and the corresponding risk rate ranged from 0.1-0.9. The nomogram model also predicted a C-index value of 0.753 (95%CI 0.726-0.781) and an area under the receiver operating characteristic curve of DN of 0.753. Internal verification of the C-index reached 0.738. The model displayed medium predictive power and could be applied in clinical practice. Conclusions: HbA1c, systolic blood pressure, Scr, GLOB, DR, and more than a junior high school education are independent risk factors of DN. Nomogram modeling can more intuitively evaluate the risk of DN in primary T2DM patients.