Fig. 1. Evolutionary outcome of advancement of optics and the phasing out of electronics, and four degrees of freedom of optical IO. (a) Pluggable optical IO; (b) on-board optical IO; (c) co-packaged optical IO; (d) three-dimensional integrated optical IO; (e) four degrees of freedom to boost optical IO bandwidth

Fig. 2. Different material platforms and corresponding research outcome. (a) Modulation speeds of different material platform^[28]; (b) lanthanum-modified lead zirconate titanate (PLZT) material^[29]; (c) SOH material; (d) OEO material+plasmonic mode^[30]; (e) barium titanate (BTO) material+plasmonic mode^[31]

Fig. 3. Research results on low-speed optical IO. (a) Schematic diagram of optical IO in optoelectronic vertical packaging; (b) Nubis2D planar optical IO chip^[49]

Fig. 4. OCS based on wavelength selection. (a) Integrated photonic chip contains AWG and SOA^[71]; (b) OCS based on WSS^[74]

Fig. 5. Spatial diffractive neural network. (a) Schematic diagram of diffraction deep neural network structure; (b) Fourier space diffraction deep neural network; (c)(d) schematic diagram of reconfigurable diffraction optoelectronic processor structure

Fig. 6. Spatial optoelectronic computing system. (a) Schematic diagram of optoelectronic hybrid computing system based on 4f system for optical convolution^[80]; (b) all-optical photon neural network based on spatial modulator and lens^[81]; (c) structural diagram of simulation iterative machine based on optoelectronic systems^[82]; (d) schematic diagram of multi-layer optoelectronic neural network^[83]

Fig. 7. Spatial optoelectronic RC system. (a) System diagram of optoelectronic hybrid RC system based on light source diffraction coupling^[84]; (b) experimental setup diagram of optoelectronic hybrid RC based on multiple scattering media^[85]; (c) schematic diagram of optoelectronic hybrid RC system using complex encoding^[86]

Fig. 8. Fiber-based optoelectronic computing system. (a) Schematic diagram of experimental setup^[96]; (b) schematic diagram of OEPO (optoelectronic parametric oscillator)-based Ising machine^[98]

Fig. 9. Fiber-based novel optoelectronic computing system. (a) Operating principle of fiber optic computing system using distributed feedback^[107]; (b) schematic diagram of photonic probabilistic processor^[108]

Fig. 10. Integrated photonic chips based on MZI array. (a) Programmable nanophotonic processor for solving Ising problems^[114]; (b) chip for complex-valued neural network with Reck scheme^115]; (c) architectures of GridNet (left) and FFTNet (right)^[116]

Fig. 11. On-chip diffractive neural networks. (a) Schematic of optical integrated diffractive neural network^[119]; (b) schematic of on-chip diffractive optical neural network^[120]

Fig. 12. Photonic chips based on non-MZI coupled waveguide network. (a) Schematic of single neuron in all-optical spiking neurosynaptic network, where red blocks represent PCM; (b) schematic of integrated photonic tensor core, where red bars represent PCM; (c) working principle of programmable photonic solver based on MRR array when solving SSP

Fig. 13. Architecture of ACCEL^[127]

Fig. 14. Analysis of bandwidth and computational capacity with a direct connection between electronic computing unit and optical computing unit (the O/E at the input end refers to an electro-optic modulation device, and the O/E at the output end refers to a photoelectric detection device)

Fig. 15. Schematic diagram of optoelectronic interconnection. (a) Optical computing unit is connected with N electronic computing units; (b) optical computing and interconnection unit is connected with N electronic computing units