Meta-models for transfer learning in source localization
In practice, nondestructive testing (NDT) procedures tend to consider experiments (and their respective models) as distinct, conducted in isolation, and associated with independent data. In contrast, this work looks to capture the interdependencies between acoustic emission (AE) experiments (as meta...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Cambridge University Press
2024-01-01
|
| Series: | Data-Centric Engineering |
| Subjects: | |
| Online Access: | https://www.cambridge.org/core/product/identifier/S2632673624000431/type/journal_article |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1841526488049909760 |
|---|---|
| author | Lawrence A. Bull Matthew R. Jones Elizabeth J. Cross Andrew Duncan Mark Girolami |
| author_facet | Lawrence A. Bull Matthew R. Jones Elizabeth J. Cross Andrew Duncan Mark Girolami |
| author_sort | Lawrence A. Bull |
| collection | DOAJ |
| description | In practice, nondestructive testing (NDT) procedures tend to consider experiments (and their respective models) as distinct, conducted in isolation, and associated with independent data. In contrast, this work looks to capture the interdependencies between acoustic emission (AE) experiments (as meta-models) and then use the resulting functions to predict the model hyperparameters for previously unobserved systems. We utilize a Bayesian multilevel approach (similar to deep Gaussian Processes) where a higher-level meta-model captures the inter-task relationships. Our key contribution is how knowledge of the experimental campaign can be encoded between tasks as well as within tasks. We present an example of AE time-of-arrival mapping for source localization, to illustrate how multilevel models naturally lend themselves to representing aggregate systems in engineering. We constrain the meta-model based on domain knowledge, then use the inter-task functions for transfer learning, predicting hyperparameters for models of previously unobserved experiments (for a specific design). |
| format | Article |
| id | doaj-art-4433e15fe51a4e82be64a4d54db50bfb |
| institution | Kabale University |
| issn | 2632-6736 |
| language | English |
| publishDate | 2024-01-01 |
| publisher | Cambridge University Press |
| record_format | Article |
| series | Data-Centric Engineering |
| spelling | doaj-art-4433e15fe51a4e82be64a4d54db50bfb2025-01-16T21:46:42ZengCambridge University PressData-Centric Engineering2632-67362024-01-01510.1017/dce.2024.43Meta-models for transfer learning in source localizationLawrence A. Bull0https://orcid.org/0000-0002-0225-5010Matthew R. Jones1https://orcid.org/0000-0001-7446-0833Elizabeth J. Cross2https://orcid.org/0000-0001-5204-1910Andrew Duncan3https://orcid.org/0000-0001-5762-164XMark Girolami4https://orcid.org/0000-0003-3008-253XSchool of Mathematics and Statistics, University of Glasgow, Glasgow, G12 8TA, UK Department of Engineering, University of Cambridge, Cambridge, CB3 0FA, UKDepartment of Engineering, University of Sheffield, Sheffield, S1 3JD, UKDepartment of Engineering, University of Sheffield, Sheffield, S1 3JD, UKDepartment of Mathematics, Imperial College London, London, SW7 2AZ, UK The Alan Turing Institute, The British Library, London, NW1 2DB, UKDepartment of Engineering, University of Cambridge, Cambridge, CB3 0FA, UK The Alan Turing Institute, The British Library, London, NW1 2DB, UKIn practice, nondestructive testing (NDT) procedures tend to consider experiments (and their respective models) as distinct, conducted in isolation, and associated with independent data. In contrast, this work looks to capture the interdependencies between acoustic emission (AE) experiments (as meta-models) and then use the resulting functions to predict the model hyperparameters for previously unobserved systems. We utilize a Bayesian multilevel approach (similar to deep Gaussian Processes) where a higher-level meta-model captures the inter-task relationships. Our key contribution is how knowledge of the experimental campaign can be encoded between tasks as well as within tasks. We present an example of AE time-of-arrival mapping for source localization, to illustrate how multilevel models naturally lend themselves to representing aggregate systems in engineering. We constrain the meta-model based on domain knowledge, then use the inter-task functions for transfer learning, predicting hyperparameters for models of previously unobserved experiments (for a specific design).https://www.cambridge.org/core/product/identifier/S2632673624000431/type/journal_articledamage localizationdeep Gaussian processesmeta-modelsmultilevel modelstransfer learning |
| spellingShingle | Lawrence A. Bull Matthew R. Jones Elizabeth J. Cross Andrew Duncan Mark Girolami Meta-models for transfer learning in source localization Data-Centric Engineering damage localization deep Gaussian processes meta-models multilevel models transfer learning |
| title | Meta-models for transfer learning in source localization |
| title_full | Meta-models for transfer learning in source localization |
| title_fullStr | Meta-models for transfer learning in source localization |
| title_full_unstemmed | Meta-models for transfer learning in source localization |
| title_short | Meta-models for transfer learning in source localization |
| title_sort | meta models for transfer learning in source localization |
| topic | damage localization deep Gaussian processes meta-models multilevel models transfer learning |
| url | https://www.cambridge.org/core/product/identifier/S2632673624000431/type/journal_article |
| work_keys_str_mv | AT lawrenceabull metamodelsfortransferlearninginsourcelocalization AT matthewrjones metamodelsfortransferlearninginsourcelocalization AT elizabethjcross metamodelsfortransferlearninginsourcelocalization AT andrewduncan metamodelsfortransferlearninginsourcelocalization AT markgirolami metamodelsfortransferlearninginsourcelocalization |