Optical networks are evolving toward ultrawide bandwidth and autonomous operation. In this scenario, it is crucial to accurately model and control optical power evolutions (OPEs) through optical amplifiers (OAs), as they directly affect the signal-to-noise ratio and fiber nonlinearities. However, a fundamental contradiction arises between the complex physical phenomena in optical transmission and the required precision in network control. Traditional theoretical methods underperform due to ideal assumptions, while data-driven approaches entail exorbitant costs associated with acquiring massive amounts of data to achieve the desired level of accuracy. In this work, we propose a Bayesian inference framework (BIF) to construct the digital twin of OAs and control OPE in a data-efficient manner. Only the informative data are collected to balance the exploration and exploitation of the data space, thus enabling efficient autonomous-driving optical networks (ADONs). Simulations and experiments demonstrate that the BIF can reduce the data size for modeling erbium-doped fiber amplifiers by 80% and Raman amplifiers by 60%. Within 30 iterations, the optimal controlling performance can be achieved to realize target signal/gain profiles in links with different types of OAs. The results show that the BIF paves the way to accurately model and control OPE for future ADONs.

As an optical processor, a diffractive deep neural network (D²NN) utilizes engineered diffractive surfaces designed through machine learning to perform all-optical information processing, completing its tasks at the speed of light propagation through thin optical layers. With sufficient degrees of freedom, D²NNs can perform arbitrary complex-valued linear transformations using spatially coherent light. Similarly, D²NNs can also perform arbitrary linear intensity transformations with spatially incoherent illumination; however, under spatially incoherent light, these transformations are nonnegative, acting on diffraction-limited optical intensity patterns at the input field of view. Here, we expand the use of spatially incoherent D²NNs to complex-valued information processing for executing arbitrary complex-valued linear transformations using spatially incoherent light. Through simulations, we show that as the number of optimized diffractive features increases beyond a threshold dictated by the multiplication of the input and output space-bandwidth products, a spatially incoherent diffractive visual processor can approximate any complex-valued linear transformation and be used for all-optical image encryption using incoherent illumination. The findings are important for the all-optical processing of information under natural light using various forms of diffractive surface-based optical processors.

Space division multiplexing (SDM) is promising to enhance capacity limits of optical networks. Among implementation options, few-mode fibres (FMFs) offer high efficiency gains in terms of integratability and throughput per volume. However, to achieve low insertion loss and low crosstalk, the beam launching should match the fiber modes precisely. We propose an all-optical data-driven technique based on multiplane light conversion (MPLC) and neural networks (NNs). By using a phase-only spatial light modulator (SLM), spatially separated input beams are transformed independently to coaxial output modes. Compared to conventional offline calculation of SLM phase masks, we employ an intelligent two-stage approach that considers knowledge of the experimental environment significantly reducing misalignment. First, a single-layer NN called Model-NN learns the beam propagation through the setup and provides a digital twin of the apparatus. Second, another single-layer NN called Actor-NN controls the model. As a result, SLM phase masks are predicted and employed in the experiment to shape an input beam to a target output. We show results on a single-passage configuration with intensity-only shaping. We achieve a correlation between experiment and network prediction of 0.65. Using programmable optical elements, our method allows the implementation of aberration correction and distortion compensation techniques, which enables secure high-capacity long-reach FMF-based communication systems by adaptive mode multiplexing devices.

Bismuth-doped fiber amplifiers offer an attractive solution for meeting continuously growing enormous demand on the bandwidth of modern communication systems. However, practical deployment of such amplifiers require massive development and optimization efforts with the numerical modeling being the core design tool. The numerical optimization of bismuth-doped fiber amplifiers is challenging due to a large number of unknown parameters in the conventional rate equations models. We propose here a new approach to develop a bismuth-doped fiber amplifier model based on a neural network purely trained with experimental data sets in E- and S-bands. This method allows a robust prediction of the amplifier operation that incorporates variations of fiber properties due to manufacturing process and any fluctuations of the amplifier characteristics. Using the proposed approach the spectral dependencies of gain and noise figure for given bi-directional pump currents and input signal powers have been obtained. The low mean (less than 0.19 dB) and standard deviation (less than 0.09 dB) of the maximum error are achieved for gain and noise figure predictions in the 1410–1490 nm spectral band.

随着大数据、云服务和人工智能的发展，网络向着规模化与复杂化方向演进，网络发生故障的概率增加，需要为每个连接请求提供专用保护，减少因网络故障造成的损失。因此，在数据中心弹性光网络中，为了提高网络的生存性，采用带宽调整方法。文章提出了一种数据中心弹性光网络专用保护优化方法。仿真结果表明，文章所提专用保护优化方法能够有效降低网络的阻塞率，提高网络资源利用率，解决了网络生存性问题。

随着光网络结构越加庞大复杂，光网络故障更易发生。网络故障发生后，由于告警间的衍生特性，根源告警会衍生出多个衍生告警，因此网管系统会收到告警风暴。由于故障和告警间关系复杂，定位网络故障，尤其是多故障，难度也急剧攀升。针对此问题，文章将擅长管理海量信息并揭示数据规律的知识图谱技术引入到光网络中，告警知识图谱蕴含了丰富的告警间关系，结合人工智能技术图神经网络（GNN）后可完成知识引导的网络故障根因自动推理，符合运维人员运维过程中的故障定位思路。进一步地，在故障定位过程中增加了网络拓扑信息，提升了知识图谱的知识维度，解除了单故障场景的限制，并在多故障定位场景中得到了较高的准确率。