Conventional meta‐atom designs rely heavily on researchers’ prior knowledge and trial‐and‐error searches using full‐wave simulations, resulting in time‐consuming and inefficient processes. Inverse design methods based on optimization algorithms, such as evolutionary algorithms, and topological optimizations, have been introduced to design metamaterials. However, none of these algorithms are general enough to fulfill multi‐objective tasks. Recently, deep learning methods represented by generative adversarial networks (GANs) have been applied to inverse design of metamaterials, which can directly generate high‐degree‐of‐freedom meta‐atoms based on S‐parameters requirements. However, the adversarial training process of GANs makes the network unstable and results in high modeling costs. This paper proposes a novel metamaterial inverse design method based on the diffusion probability theory. By learning the Markov process that transforms the original structure into a Gaussian distribution, the proposed method can gradually remove the noise starting from the Gaussian distribution and generate new high‐degree‐of‐freedom meta‐atoms that meet S‐parameters conditions, which avoids the model instability introduced by the adversarial training process of GANs and ensures more accurate and high‐quality generation results. Experiments have proven that our method is superior to representative methods of GANs in terms of model convergence speed, generation accuracy, and quality.

Advanced digital signal processing (DSP) is viewed as an inevitable technology to deal with nonlinear distortions in cost-sensitive optical interconnection systems. Classic methods such as equalization based on Volterra series turn out to be quite effective solutions to compensate signal impairments. In practical implementations, a great deal of efforts focus on reducing the computational complexity. Several model selection and dimensionality reduction approaches have been proposed to simplify the structure, which employ backward strategies. However, the threshold settings for these pruning approaches are environment-sensitive. In this paper, we propose a least angle regression-based nonlinear equalization method (LaNLE). We extend the non-intrusive least angle regression technique to generalized linear models. Through an incremental process by stagewise feature selection and weight fitting, we obtain the low-complexity nonlinear structure with significant feature basis. Experimental verification is performed based on 120 Gbps PAM8-modulated 850 nm VCSEL plus MMF optical link. The results show that, compared to fully-connected Volterra nonlinear equalizer structure (VNLE), up to 60.4% and 70.1% reduction in complexity can be achieved with similar BER performance in back-to-back and 100 m transmission scenarios, respectively. In addition, the LaNLE method has a complete path of segmented linear results where the performance shows stable growth with increasing feature basis. We experimentally demonstrate 225 Gbps per lane PAM-8 signal transmission over 100 m MMF in the simple and cost-effective implementation manner based on mass-production 850 nm VCSEL and proposed LaNLE method.

Reconfigurable intelligent surfaces (RIS) have been considered as a promising solution to enhance the spectrum and energy efficiency of the 5G+ and future 6G wireless communication systems. The performance of RIS will become the key metric of these communication systems. In this paper, we proposed a wideband Lorenz resonance-based metasurface reconfigurable reflectarray (MSRRA) realization scheme with low energy consumption targeted at the center frequency of 28 GHz. A compact voltage bias network for the varactor diodes is carefully designed to reduce losses in RF current and the influence of the bias circuit on the radiating element effectively. An equivalent circuit model for the MSRRA unit cell is also introduced to predict the properties of the MSRRA system, which can be used to optimize the MSRRA design efficiently. In the experimental tests, the proposed MSRRA system can be optimized to cover a dynamic reflection phase range of over 300° with a bandwidth of 3.83 GHz, which is consistent with the simulation results. The measured single-scattering beam bandwidth is 1.85 GHz at the center frequency of 28 GHz, which can fully cover the whole n257 channel of 5G NR. The proposed continuous tunable MSRRA can support 5G+ and 6G indoor, short-range links, and outdoor point-to-point communications.

Miniaturized spectrometers employing photonic crystal cavity arrays in conjunction with computational reconstruction have gained attention as effective tools for spectral analysis. Nevertheless, achieving an optimal balance among spectral resolution, detection range, and device compactness remains challenging, particularly when complex nonlinear mappings, inter-pattern correlations, and noise interference are involved. In this work, we present ESTspecNet, a deep learning framework that integrates EfficientNet, the Swin Transformer, and spatial-channel attention mechanisms to improve spectral reconstruction accuracy. We reconstructed near-infrared spectra over an 80 nm range using a 144-unit photonic crystal cavity array, and achieved a single-peak resolution of $0.47$ nm and a double-peak resolution of $0.7$ nm. Compared to conventional methods, the proposed model demonstrates superior performance in both wide-range spectral reconstruction and fine-resolution tasks, thus highlighting its ability to effectively capture intricate spectral features and long-range dependencies, thereby advancing the reconstruction capabilities of miniaturized spectrometers.

In ITU-T standards, auxiliary management and control channels (AMCCs), as defined, facilitate the rapid deployment and efficient management of wavelength division multiplexing passive optical network (WDM-PON) systems. The super-imposition of an AMCC introduces additional interference to a PON signal, resulting in the degradation of the performance of the overall transmission. In prior research, we proposed employing a Gaussian mixture model (GMM) to fit a baseband-modulated AMCC signal. Following the analysis of the interference model and the distribution characteristics of received signal errors, we propose a combined optimization method for a transmitter and receiver in this paper. This method, grounded in probabilistic shaping (PS) techniques, optimizes the probability distribution of the transmitted signal based on the AMCC interference model, with the objective of reducing the error rate in PON signal transmission. We have validated this approach within a 50G-PON experimental system by utilizing PAM4 modulation. The experimental results demonstrate the effectiveness of this method for mitigating the impact of baseband-modulated AMCC, thereby reducing the error rate in PON signal transmission. The approach presented in this paper can further minimize the performance degradation introduced by baseband-modulated AMCC in WDM-PON systems, enhancing the efficiency of WDM-PON deployment.