Scholarlink

<p>In this paper, we design a deep convolutional generative adversarial network model for a programmable coding metasurface inverse design, which effectively overcomes the limitations of deep learning in metasurface inverse design by incorporating a self-attention mechanism. Rather than just a mechanical replication of a predefined pattern within the training set, the methodology provides in-depth learning of the inverse mapping relationship that exists between the metasurface structure and its electromagnetic response. The output structure of the network model not only conforms to the basic design rules of the metasurface, but also can effectively reverse derive and generate novel and accurate design solutions outside the training set, as well as achieving good design performance.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Inverse Design of Programmable Coding Metasurface Based on Deep Convolutional Generative Adversarial Networks

  • 期刊论文

  • 发表日期：

    2025年9月2日

查看全文
  • Shanhui Liu,
  • Nengwu Hu,
  • Peng Xu,
  • Qiuhuang Chen,
  • Jinghui Fang,
  • Jun Lou,
  • Ying Tian,
  • Xufeng Jing

摘要

In this paper, we design a deep convolutional generative adversarial network model for a programmable coding metasurface inverse design, which effectively overcomes the limitations of deep learning in metasurface inverse design by incorporating a self-attention mechanism. Rather than just a mechanical replication of a predefined pattern within the training set, the methodology provides in-depth learning of the inverse mapping relationship that exists between the metasurface structure and its electromagnetic response. The output structure of the network model not only conforms to the basic design rules of the metasurface, but also can effectively reverse derive and generate novel and accurate design solutions outside the training set, as well as achieving good design performance.