Towards General Neural Surrogate Solvers with Specialized Neural Accelerators
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866916557448282112 |
|---|---|
| author | Mao, Chenkai Lupoiu, Robert Dai, Tianxiang Chen, Mingkun Fan, Jonathan A. |
| author_facet | Mao, Chenkai Lupoiu, Robert Dai, Tianxiang Chen, Mingkun Fan, Jonathan A. |
| contents | Surrogate neural network-based partial differential equation (PDE) solvers have the potential to solve PDEs in an accelerated manner, but they are largely limited to systems featuring fixed domain sizes, geometric layouts, and boundary conditions. We propose Specialized Neural Accelerator-Powered Domain Decomposition Methods (SNAP-DDM), a DDM-based approach to PDE solving in which subdomain problems containing arbitrary boundary conditions and geometric parameters are accurately solved using an ensemble of specialized neural operators. We tailor SNAP-DDM to 2D electromagnetics and fluidic flow problems and show how innovations in network architecture and loss function engineering can produce specialized surrogate subdomain solvers with near unity accuracy. We utilize these solvers with standard DDM algorithms to accurately solve freeform electromagnetics and fluids problems featuring a wide range of domain sizes. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2405_02351 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Towards General Neural Surrogate Solvers with Specialized Neural Accelerators Mao, Chenkai Lupoiu, Robert Dai, Tianxiang Chen, Mingkun Fan, Jonathan A. Machine Learning Artificial Intelligence Distributed, Parallel, and Cluster Computing Optics Surrogate neural network-based partial differential equation (PDE) solvers have the potential to solve PDEs in an accelerated manner, but they are largely limited to systems featuring fixed domain sizes, geometric layouts, and boundary conditions. We propose Specialized Neural Accelerator-Powered Domain Decomposition Methods (SNAP-DDM), a DDM-based approach to PDE solving in which subdomain problems containing arbitrary boundary conditions and geometric parameters are accurately solved using an ensemble of specialized neural operators. We tailor SNAP-DDM to 2D electromagnetics and fluidic flow problems and show how innovations in network architecture and loss function engineering can produce specialized surrogate subdomain solvers with near unity accuracy. We utilize these solvers with standard DDM algorithms to accurately solve freeform electromagnetics and fluids problems featuring a wide range of domain sizes. |
| title | Towards General Neural Surrogate Solvers with Specialized Neural Accelerators |
| topic | Machine Learning Artificial Intelligence Distributed, Parallel, and Cluster Computing Optics |
| url | https://arxiv.org/abs/2405.02351 |