Adaptive entropy loading via direct activation for an optical frequency comb-based few-mode transmission system

We demonstrate a multi-dimensional entropy loading mechanism in an optical frequency comb-based few-mode transmission system. The frequency comb functions as a compact and energy-efficient multi-carrier laser source; however, performance discrepancies arise across wavelength channels and spatial modes due to power variation among comb lines, intermodal crosstalk, and environmental fluctuations. To address these impairments, we employ a neuronal direct activation structure that enables multi-dimensional entropy loading for adaptive reallocation of the source entropy. Our unique solution is a model-free, neuromorphic-inspired mapping framework that utilizes tanh activation and max-pooling to achieve instantaneous entropy reallocation. This approach provides iterative-free dynamic entropy adjustment, ensuring efficient channel utilization and instantaneous adaptability to varying network conditions. Experimental results verify that the proposed scheme effectively mitigates non-uniform channel performance, achieving a 4.82% capacity enhancement, corresponding to a data rate increase of 31.7 Gbit/s. Our approach extends the entropy loading to multi-dimensional coherent systems, offering a highly scalable and robust solution for next-generation data center interconnects.