IP Library › Granted Patent US 12,361,690
Granted Patent B2
US 12,361,690 · App. 17/931,731 · Granted Jul 15, 2025

Random sampling consensus federated semi-supervised learning

Inventor: Xiaomeng Li (Hong Kong, CN)
Assignee: The Hong Kong University of Science and Technology
G06V10/7753G06N20/00G06V10/95
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,361,690
App. No.
17/931,731
Granted
Jul 15, 2025
Kind
B2
Abstract

A method and systems for random sampling consensus federated (RSCFed) learning in non-IID settings are provided. The method includes randomly sampling local clients, assigning a current global model to the randomly sampled local clients for initialization at beginning of a synchronization round, conducting local training on the randomly sampled local clients, collecting local models from the randomly sampled local clients and executing distance-reweighted model aggregation (DMA) on the collected local models to obtain a sub-consensus model, repeating above steps multiple times to obtain a set of sub-consensus models, and aggregating a new model based on the sub-consensus models to be next global model.

Claims (38)

1. A method of random sampling consensus federated (RSCFed) learning, comprising:

randomly sampling local clients;

assigning a current global model to the randomly sampled local clients for initialization at beginning of a synchronization round;

conducting local training on the randomly sampled local clients;

collecting local models from the randomly sampled local clients and executing distance-reweighted model aggregation (DMA) on the collected local models to obtain a sub-consensus model;

repeating the foregoing steps multiple times to obtain a set of sub-consensus models; and

aggregating a new model based on the sub-consensus models to be a next global model.

2. The method of claim 1 , wherein the local clients comprise labeled local clients having labeled local data and unlabeled local clients having unlabeled local data.

3. The method of claim 1 , wherein the step of assigning a current global model to the randomly sampled clients for initialization comprises initializing the local models with the current global model for performing local training on the randomly sampled clients.

4. The method of claim 2 , wherein the step of conducting local training comprises conducting standard supervised and unsupervised training on the labeled and unlabeled local clients, respectively.

5. The method of claim 1 , wherein the local training on the labeled local clients is conducted with a main objective, cross-entropy loss, L CE , defined by Equation:

L CE =−y i log( ŷ i );

where ŷ i is a prediction of the randomly sampling local clients from a corresponding local model.

6. The method of claim 2 , wherein the local training on the unlabeled local clients is conducted by a mean-teacher-based consistency regularization framework and regarding a student model as the local model.

7. The method of claim 1 , wherein the distance-reweighted model aggregation (DMA) is configured to dynamically adjust weights of the collected models.

8. The method of claim 2 , wherein during the local training on the unlabeled local clients, after predictions from the student model and the teacher model are generated, a sharpening method is configured to increase a temperature of the predictions of the teacher model.

9. The method of claim 8 , wherein when the local training is completed, the student model is provided as the local model for a corresponding unlabeled local client.

10. The method of claim 1 , wherein the executing distance-reweighted model aggregation (DMA) comprises:

computing an intra-subset averaged model for each subset;

scaling weight for the local client in each subset; and

normalizing the intra-subset model weight into a range of [0, 1].

11. A system for performing random sampling consensus federated (RSCFed) learning, comprising:

a federal server coupled to a plurality of local clients through a communication network;

wherein the federal server is configured to randomly sample the plurality of local clients; assign a current global model to the randomly sampled local clients for initialization at beginning of a synchronization round; conduct local training on the randomly sampled local clients; collect local models from the randomly sampled local clients; execute distance-reweighted model aggregation (DMA) on the collected local models to obtain a sub-consensus model; repeat the foregoing steps multiple times to obtain a set of sub-consensus models; and aggregate a new model based on the sub-consensus models to be a next global model.

12. The system of claim 11 , wherein the local clients comprise labeled clients having labeled local data and unlabeled clients having unlabeled local data.

13. The system of claim 11 , wherein assigning a current global model to the randomly sampled clients for initialization comprises initializing local models with the current global model for performing local training on the randomly sampled clients.

14. The system of claim 12 , wherein the local training comprises conducting standard supervised and unsupervised training on the labeled and unlabeled local clients, respectively.

15. The system of claim 12 , wherein the local training on the labeled local clients is conducted with a main objective, cross-entropy loss, LCE, defined by Equation:

L CE =−y i log( ŷ i );

where ŷ i is a prediction of the randomly sampled local clients from a corresponding local model.

16. The system of claim 12 , wherein the local training on the unlabeled local clients is conducted by a mean-teacher-based consistency regularization framework and regarding a student model as the local model.

17. The system of claim 11 , wherein the distance-reweighted model aggregation (DMA) is configured to dynamically adjust weights of the collected models.

18. The system of claim 12 , wherein during the local training on the unlabeled local clients, after predictions from the student model and the teacher model are generated, a sharpening method is configured to increase a temperature of the predictions of the teacher model.

19. The system of claim 18 , wherein when the local training is completed, the student model is provided as the local model for a corresponding unlabeled local client.

20. The system of claim 11 , wherein the distance-reweighted model aggregation (DMA) is performed by:

computing an intra-subset averaged model for each subset;

scaling weight for a local client in each subset; and

normalizing the intra-subset model weight into a range of [0, 1].

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Nov 2, 2022
From: LI, XIAOMENG
To: THE HONG KONG UNIVERSITY OF SCIENCE AND TECHNOLOGY
Reel/Frame 061629/0350 →
Continuity (2)
Provisional Application 63287093 · Dec 8, 2021
Related Publication 20230177812A1 · Jun 8, 2023
References Cited (50)
US 20180336486A1 · Chu · 2018 [cited by examiner]
US 20200394552A1 · Ganapavarapu · 2020 [cited by examiner]
US 20210073677A1 · Peterson · 2021 [cited by examiner]
US 20210150269A1 · Choudhury · 2021 [cited by examiner]
US 20210312336A1 · Sinn · 2021 [cited by examiner]
US 20220012637A1 · Rezazadegan Tavakoli · 2022 [cited by examiner]
US 20220058507A1 · El-Khamy · 2022 [cited by examiner]
US 20220129706A1 · Vivona · 2022 [cited by examiner]
US 20240070530A1 · Amid · 2024 [cited by examiner]
US 20240086718A1 · Malaviya · 2024 [cited by examiner]
CN 113691594 · 2021 [cited by examiner]
Liang, Xiaoxiao, et al. “Rscfed: Random sampling consensus federated semi-supervised learning.” Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2022. (Year: 2022). [cited by examiner]
Lin, Haowen, et al. “Semifed: Semi-supervised federated learning with consistency and pseudo-labeling.” arXiv preprint arXiv: 2108.09412 (2021). (Year: 2021). [cited by examiner]
Yang, Dong, et al. “Federated semi-supervised learning for COVID region segmentation in chest CT using multi-national data from China, Italy, Japan.” Medical image analysis 70 (2021): 101992. (Year: 2021). [cited by examiner]
Hegedüs, István, Gábor Danner, and Márk Jelasity. “Decentralized learning works: An empirical comparison of gossip learning and federated learning.” Journal of Parallel and Distributed Computing 148 (2021): 109-124. (Ye… [cited by examiner]
Berthelot, D., et al., “MixMatch: A Holistic Approach to Semi-Supervised Learning,” arXiv preprint, arXiv:1905.02249v2 [cs.LG], Oct. 23, 2019, pp. 1-14. [cited by applicant]
Chen, H.-Y., et al., “Fedbe: Making Bayesian Model Ensemble Applicable to Federated Learning,” arXiv preprint, arXiv:2009.01974v4 [cs.LG], Oct. 10, 2021, pp. 1-21. [cited by applicant]
Chen, X., et al., “Semi-Supervised Semantic Segmentation with Cross Pseudo Supervision,” arXiv preprint, arXiv:2106.01226v2 [cs.CV], Jun. 4, 2021, pp. 1-10. [cited by applicant]
Cui, W., et al., “Semi-Supervised Brain Lesion Segmentation with an Adapted Mean Teacher Model,” arXiv preprint, arXiv:1903.01248v1 [cs.CV], Mar. 4, 2019, pp. 1-12. [cited by applicant]
Fischler, M. A., et al., “Random Sample Consensus: A Paradigm for Model Fitting with Applications to Image Analysis and Automated Cartography,” Communications of the ACM, Jun. 1981, 24(6):381-395. [cited by applicant]
He, K., et al., “Deep Residual Learning for Image Recognition,” arXiv preprint, arXiv:1512.03385v1 [cs.CV], Dec. 10, 2015, pp. 1-12. [cited by applicant]
Hu, Z., et al., “SimPLE: Similar Pseudo Label Exploitation for Semi-Supervised Classification,” arXiv preprint, arXiv:2103.16725v2 [cs.CV], Jun. 29, 2022, pp. 1-13. [cited by applicant]
Jeong, W., et al., “Federated Semi-Supervised Learning With Inter-Client Consistency & Disjoint Learning,” arXiv preprint, arXiv:2006.12097v3 [cs.LG], Mar. 29, 2021 pp. 1-15. [cited by applicant]
Kairouz, P., et al., “Advances and Open Problems in Federated Learning,” arXiv preprint, arXiv: 1912.04977v3 [cs.LG], Mar. 9, 2021, pp. 1-121. [cited by applicant]
Kaissis, G.A., et al., “Secure, privacy-preserving and federated machine learning in medical imaging,” Nature Machine Intelligence, Jun. 2020, 2:305-311. [cited by applicant]
Kang, Y., et al., “FedCVT: Semi-Supervised Vertical Federated Learning with Cross-View Training,” J. ACM, Jan. 2022, 1(1):1-17. [cited by applicant]
Karimireddy, S.P., et al., “SCAFFOLD: Stochastic Controlled Averaging for Federated Learning,” arXiv preprint, arXiv:1910.06378v4 [cs.LG], Apr. 9, 2021, pp. 1-41. [cited by applicant]
Konecny, J., et al., “Federated Learning: Strategies for Improving Communication Efficiency,” arXiv preprint, arXiv:1610.05492v2 [cs.LG], Oct. 30, 2017, pp. 1-10. [cited by applicant]
Kumar, R., et al., “Blockchain-Federated-Learning and Deep Learning Models for COVID-19 detection using CT Imaging,” arXiv preprint, arXiv:2007.06537v2 [eess.IV], Dec. 8, 2020, pp. 1-16. [cited by applicant]
Li, Q., et al., “Federated Learning on Non-IID Data Silos: An Experimental Study,” arXiv preprint, arXiv:2102.02079v4 [cs.LG], Oct. 28, 2021, pp. 1-20. [cited by applicant]
Li, T., et al., “Federated Optimization in Heterogeneous Networks,” arXiv preprint, arXiv:1812.06127v5 [cs.LG], Apr. 21, 2020, pp. 1-22. [cited by applicant]
Li, X., et al., “Transformation-consistent Self-ensembling Model for Semi-supervised Medical Image Segmentation,” arXiv preprint, arXiv:1903.00348v3 [cs.CV], May 8, 2020, pp. 1-12. [cited by applicant]
Lin, H., et al., “SemiFed: Semi-supervised Federated Learning with Consistency and Pseudo-Labeling,” arXiv preprint, arXiv:2108.09412v1 [cs.LG], Aug. 21, 2021, pp. 1-10. [cited by applicant]
Liu, Q., et al., “FedDG: Federated Domain Generalization on Medical Image Segmentation via Episodic Learning in Continuous Frequency Space,” arXiv preprint, arXiv:2103.06030v1 [cs.CV], Mar. 10, 2021, pp. 1-9. [cited by applicant]
Liu, Q., et al., “Federated Semi-supervised Medical Image Classification via Inter-client Relation Matching,” arXiv preprint, arXiv:2106.08600v1 [cs.CV], Jun. 16, 2021, pp. 1-11. [cited by applicant]
Liu, Y., et al., “FedVision: An Online Visual Object Detection Platform Powered by Federated Learning,” arXiv preprint, arXiv:2001.06202v1 [cs.LG], Jan. 17, 2020 pp. 1-8. [cited by applicant]
McMahan, H.B., et al., “Communication-Efficient Learning of Deep Networks from Decentralized Data,” arXiv preprint, arXiv:1602.05629v4 [cs.LG], Jan. 26, 2023, pp. 1-11. [cited by applicant]
Sohn, K., et al., “FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence,” 34th Conference on Neural Information Processing Systems, 2020, pp. 1-21. [cited by applicant]
Tarvainen, A., et al., “Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results,” arXiv preprint, arXiv:1703.01780v6 [cs.NE], Apr. 16, 2018, pp. 1-16. [cited by applicant]
Wang, H., et al., “Federated learning with matched averaging,” arXiv preprint, arXiv:2002.06440v1 [cs.LG], Feb. 15, 2020, pp. 1-16. [cited by applicant]
Wang, J., et al., “Tackling the Objective Inconsistency Problem in Heterogeneous Federated Optimization,” arXiv preprint, arXiv:2007.07481v1 [cs.LG], Jul. 15, 2020, pp. 1-34. [cited by applicant]
Yang, D., et al., “Federated Semi-Supervised Learning for COVID Region Segmentation in Chest CT using Multi-National Data from China, Italy, Japan,” Medical Image Analysis, 2020, pp. 1-19. [cited by applicant]
Yang, Q., et al., “Federated Machine Learning: Concept and Applications,” ACM Trans. Intell. Syst. Technol., Feb. 2019, 10(2):1-19. [cited by applicant]
Yin, H., et al., “See through Gradients: Image Batch Recovery via GradInversion,” arXiv preprint, arXiv:2104.07586v1 [cs.LG], Apr. 15, 2021, pp. 1-13. [cited by applicant]
Yoon, T., et al., “Fedmix: Approximation of Mixup Under Mean Augmented Federated Learning,” arXiv preprint, arXiv:2107.00233v1 [cs.LG], Jul. 1, 2021, pp. 1-19. [cited by applicant]
Yu, L., et al., “Uncertainty-aware Self-ensembling Model for Semi-supervised 3D Left Atrium Segmentation,” arXiv preprint, arXiv:1907.07034v1 [cs.CV], Jul. 16, 2019, pp. 1-9. [cited by applicant]
Yurochkin, M., et al., “Bayesian Nonparametric Federated Learning of Neural Networks,” arXiv preprint, arXiv:1905.12022v1 [stat.ML], May 28, 2019, pp. 1-15. [cited by applicant]
Zhang, M., et al., “Personalized Federated Learning With First Order Model Optimization,” arXiv preprint, arXiv:2012.08565v4 [cs.LG], Mar. 26, 2021, pp. 1-17. [cited by applicant]
Zhang, Z., et al., “Benchmarking Semi-Supervised Federated Learning,” arXiv preprint, arXiv:2008.11364v1 [cs.LG], Aug. 26, 2020, pp. 1-19. [cited by applicant]
Zou, Y., et al., “Pseudoseg: Designing Pseudo Labels for Semantic Segmentation,” ICLR 2021, 2021, pp. 1-18. [cited by applicant]