Institutional Repository of Radio Astronomy Research Laboratory
| A multi-fidelity transfer learning strategy based on multi-channel fusion | |
Zhang, Zihan1; Ye, Qian3; Yang, Dejin2,3; Wang, Na4 ; Meng, Guoxiang1
| |
| 2024-06-01 | |
| Source Publication | JOURNAL OF COMPUTATIONAL PHYSICS
 |
| ISSN | 0021-9991 |
| Volume | 506Pages:112952 |
| Contribution Rank | 4 |
| Abstract | Multi -fidelity strategies leverage a large amount of low -fidelity data combined with a smaller set of high-fidelity data, thereby achieving satisfactory results at a reasonable cost. In our research, we introduce an innovative multi -fidelity strategy that integrates the concepts of multi -fidelity data fusion and transfer learning. In the proposed framework, we incorporate auto -encoders and a multi -channel transfer learning strategy, enabling the network model to comprehend the relationship between the low -fidelity and high-fidelity models in both explicit and implicit manners. This approach not only enhances prediction accuracy but also mitigates issues such as overfitting and negative transfer, which may arise in scenarios with sparse samples. Additionally, Bayesian optimization is employed for effective hyperparameter selection. To evaluate and analyze the performance of our proposed method, we present a series of benchmark test cases. Furthermore, we also show the application of the proposed method to engineering problems. Firstly, we consider a parametrized partial differential equation problem, where high-fidelity and low -fidelity data are obtained using exact methods and simplified algorithms, respectively. Subsequently, we extend this strategy to convolutional neural network architectures, specifically addressing a pressure Poisson equation problem. We also explore the effect of the reliability of the low -fidelity data and the number of high-fidelity data on the results. The results show that the proposed method exhibits low requirements in terms of both the reliability of the low -fidelity data and the number of high-fidelity data while maintaining satisfactory accuracy metrics. |
| Keyword | Surrogate model Deep neural networks Multi-fidelity Data fusion |
| DOI | 10.1016/j.jcp.2024.112952 |
| Indexed By | SCI |
| Language | 英语 |
| WOS Keyword | OPTIMIZATION ; SURFACES |
| Funding Project | National Key Basic Research and Development Program of China[2021YFC2203501] ; National Natural Science Foundation of China[U1931137] |
| WOS Research Area | Computer Science ; Physics |
| WOS Subject | Computer Science, Interdisciplinary Applications ; Physics, Mathematical |
| WOS ID | WOS:001221359200001 |
| Publisher | ACADEMIC PRESS INC ELSEVIER SCIENCE |
| Funding Organization | National Key Basic Research and Development Program of China ; National Natural Science Foundation of China |
| Citation statistics | |
| Document Type | 期刊论文 |
| Identifier | http://ir.xao.ac.cn/handle/45760611-7/6620 |
| Collection | 射电天文研究室_脉冲星研究团组 |
| Corresponding Author | Ye, Qian |
| Affiliation | 1.Shanghai Jiao Tong Univ, Sch Mech Engn, Shanghai 200240, Peoples R China 2.Chongqing Jiaotong Univ, Sch Civil Engn, Chongqing 400074, Peoples R China 3.Chinese Acad Sci, Shanghai Astron Observ, Shanghai 200030, Peoples R China 4.Chinese Acad Sci, Xinjiang Observ, Xinjiang 830011, Peoples R China |
| Recommended Citation GB/T 7714 | Zhang, Zihan,Ye, Qian,Yang, Dejin,et al. A multi-fidelity transfer learning strategy based on multi-channel fusion[J]. JOURNAL OF COMPUTATIONAL PHYSICS,2024,506:112952. |
| APA | Zhang, Zihan,Ye, Qian,Yang, Dejin,Wang, Na,&Meng, Guoxiang.(2024).A multi-fidelity transfer learning strategy based on multi-channel fusion.JOURNAL OF COMPUTATIONAL PHYSICS,506,112952. |
| MLA | Zhang, Zihan,et al."A multi-fidelity transfer learning strategy based on multi-channel fusion".JOURNAL OF COMPUTATIONAL PHYSICS 506(2024):112952. |
| Files in This Item: | ||||||
| File Name/Size | DocType | Version | Access | License | ||
| Zhang-2024-A multi-f（3852KB） | 期刊论文 | 出版稿 | 开放获取 | CC BY-NC-SA | View Application Full Text | |
Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.
Edit Comment