Transformers are very powerful, and also very complex. They use a dense feedforward network as a sub-neural net inside the encoder and decoder components. They also demand considerable computing ...

Zhu RC, Wang JF, Qiu TS, Yang DK, Feng B et al. Direct field-to-pattern monolithic design of holographic metasurface via residual encoder-decoder convolutional neural network. Opto-Electron Adv 6 ...