Multipass cell design with the random walk and gradient descent optimization algorithms

An automated approach is presented for optimizing the multipass cell (MPC) design with dense patterns in this paper. First, a strategy based on the random walk (RW) algorithm is implemented for global exploration to determine the parameters of the target MPC configuration and accelerate the design p...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Applied physics. B, Lasers and optics Ročník 127; číslo 9
Hlavní autoři: Kong, Rong, Liu, Peng, Zhou, Xin
Médium: Journal Article
Jazyk:angličtina
Vydáno: Berlin/Heidelberg Springer Berlin Heidelberg 01.09.2021
Springer Nature B.V
Témata:
Algorithms
Applied physics
Clustering
Configuration management
Design optimization
Engineering
Identification methods
Lasers
Optical Devices
Optics
Photonics
Physical Chemistry
Physics
Physics and Astronomy
Quantum Optics
Random walk
Regular Paper
ISSN:0946-2171, 1432-0649
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:An automated approach is presented for optimizing the multipass cell (MPC) design with dense patterns in this paper. First, a strategy based on the random walk (RW) algorithm is implemented for global exploration to determine the parameters of the target MPC configuration and accelerate the design process. Second, the gradient descent (GD) algorithm is performed for local exploitation to optimize the re-entrant condition in a fast and automatic way. In addition, we apply the clustering method to identify the desired spot patterns with specific properties automatically. Finally, the proposed algorithms are tested in the optimization of two types of densely patterned MPC under the re-entrant condition. The results presented in this paper clearly show that the proposed approach is effective and efficient in optimizing the MPC design automatically and can be further utilized in more complex optical configurations.
Bibliografie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:0946-2171
1432-0649
DOI:10.1007/s00340-021-07679-6