.png){kind=icon}
.png){kind=icon}
Return
Machine-learning-driven accelerated design-method for meta-devices
.png){kind=icon}
PRE
AI
DOI：10.1016/j.mtcomm.2023.106951.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Metasurface-based solar absorbers are widely acknowledged in green energy applications. The traditional roadmap for designing meta-absorbers relies on hefty trial-and-error computations for reaching design goals. In contrast, emerging machine-learning (ML) techniques can make such designs faster and more efficient, while conserving computational resources. ML models, i.e. Decision Tree (DT) and Random Forest (RF) Regressors are demonstrated in this work to design a variety of meta-absorbers. The models have been trained to generate desired electromagnetic spectrum, shapes, and geometries of meta-atoms during forward and inverse configurations, respectively. The MSE of DT and RF are 8.61 x 10-10 and 1.56 x 10-2 for forward, while 4.42 x 10-2 and 1.35 x 10-1 for inverse models, respectively.
Keywords:
Metasurface
Solar absorber
Refractory materials
Machine learning
Regression analysis
Regressor
Decision tree
Random forest
Journal
IF:
4.5
Papers:
1.5W
Citations:
3.7W