Accelerating shape optimization by deep neural networks with on-the-fly determined architecture Journal: Elsevier Journal
Lucie Kubíčková, Ondřej Gebouský, Jan Haidl, Martin Isoz · 2025
Read Accelerating shape optimization by deep neural networks with on-the-fly determined architecture Journal: Elsevier Journal by Lucie Kubíčková, Ondřej Gebouský, Jan Haidl, Martin Isoz (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2512-03555.txt.
Description
Shape optimization can need thousands of costly simulations. Lucie Kubíčková, Ondřej Gebouský, Jan Haidl, and Martin Isoz speed it up in this 2025 work. A global search over component shapes is infeasible when each test is expensive.
Their method pairs multi-objective evolutionary algorithms with deep neural networks. The network's architecture is set on the fly during the search. Learned surrogates replace most simulations, making global search viable. Faster design cycles help engineers in aerospace, automotive, and energy fields. On-the-fly architecture search avoids fixing the surrogate before the problem's shape is known, letting the network grow with the design space.
Catalog record
| CREATOR | Lucie Kubíčková, Ondřej Gebouský, Jan Haidl, Martin Isoz |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 82 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2512-03555