Fig. 1. Geometric structure of coding unit cell^[78]. (a) Top view; (b) cross-sectional view

Fig. 2. Flow chart of joint optimization by BPSO algorithm together with CST simulation software^[78]

Fig. 3. Models, reflection phases and amplitudes of 1 bit coding unit cells^[78]. (a) Geometric diagram of first coding unit cell; (b) geometric diagram of second coding unit cell; (c) reflection phases of 1 bit coding unit cells and their phase difference; (d) reflection amplitudes of 1 bit coding unit cells

Fig. 4. Structural diagram of coding unit cell^[82]

Fig. 5. Flow chart of machine learning design for 1 bit coding unit with phase difference in orthogonal directions^[82]

Fig. 6. Comparison of non-iterative deep learning method and conventional optimization method^[84]

Fig. 7. Schematic of layered optical metamaterial^[85]. (a) Thin films composed of multiple layers of SiO₂ and Si₃N₄; (b)(c) two 6-layered thin film structures with similar transmission spectra and their corresponding spectra

Fig. 8. Tandem network composed of inverse design network and forward modeling network^[85]

Fig. 9. Schematic of GAN-based inverse design method^[86]

Fig. 10. All-optical diffractive neural networks^[90]. (a) Schematic of network structure; (b) schematic of 3D printing structure

Fig. 11. Imaging principle and imaging results^[77]. (a) Principle diagram of imaging algorithm; (b) four images of testing person; (c) reconstructed images corresponding to Fig.11(b) using principal component analysis method; (d) reconstructed images corresponding to Fig.11(b) using random radiation pattern

Fig. 12. Schematic of imaging and recognition system^[95]. (a) Collection and construction of imaging data; (b) flow chart of recognition system

Fig. 13. Structural diagram of intelligence imaging system with joint optimization^[96]

Fig. 14. Principle diagram and application of smart metasurface^[97]; (a) Application scenario of satellite communication; (b) components of smart metasurface