Dynamic inverse design of broadband metasurfaces with synthetical neural networks

HIGHLIGHTS

• What: To address the lack of systematic compositionality, the authors explore a modular approach to metasurface inverse design with neural_networks that inherits knowledge from neural_networks previously trained for the inverse design of different segments of which the new target metasurface is composed. The authors seek to overcome this limitation with a synthetical neural_network. The authors evaluate the loss function using the mean square error (MSE) metric, in line with established practices in deep learning, and the authors train the encoder-decoder network using a standard gradient descent algorithm44. The work demonstrates great adaptability and scalability and, importantly . . .
  

   

  

  If you want to have access to all the content you need to log in!

  Thanks :)

  Username or Email

  Password

  Remember me

  Lost your password?

  If you don't have an account, you can create one here.