What is it about?
In our work on predicting MIM metasurfaces, we have achieved remarkably low error values. The results of this work contribute to the development of the field of electromagnetic high-accuracy computing in artificial intelligence. The core of our paper is proposing a neural network that is highly applicable for predicting MIM structures with high accuracy. Predicting the performance of micro and nano structures faster and more efficiently is also an important part of electromagnetic computing.
Featured Image
Photo by Ali Yılmaz on Unsplash
Why is it important?
1. Traditional electromagnetic computing software requires longer computation time for high-precision and large-sized micro/nanostructures on today's high-performance computers. 2. The next generation of quantum computers has significant limitations in computing tasks and cannot guarantee complete computation of existing electromagnetic computing tasks. 3. Neural networks require very little computer resources for prediction after training, and are considered a very fast solution. However, there are currently significant issues with prediction accuracy.
Perspectives
High-accuracy prediction in electromagnetic computing is an important problem. Thus far, numerous studies have shown the deep learning can help extremely fast design nano-photonics devices. Compared to the accuracy of traditional electromagnetic calculation software, it still was a challenges.If we can solve this problem, it will greatly improve the calculation accuracy of the next generation of artificial intelligence electromagnetic calculation software.
Kaizhu Liu
Dalian University of Technology
Read the Original
This page is a summary of: High efficiency design of metal–insulator–metal metasurface by ResNets-10, Applied Physics Letters, November 2023, American Institute of Physics,
DOI: 10.1063/5.0159954.
You can read the full text:
Contributors
The following have contributed to this page
