Enhancing Generalization in Neural Channel Model for Optical Fiber WDM Transmission through Learned Encoding of System Parameters

We propose a learned encoding method to enhance neural channel model generalization by integrating system parameters as side information. This approach achieves large-scale generalization, encompassing optical fiber transmission launch power and distance.