Fiber transmission model with parameterized inputs based on generative pre-trained physics-informed neural networks

In this manuscript, a principle-driven fiber transmission model for short-distance transmission with parameterized inputs is put forward. By taking into account the previously proposed principle-driven fiber model as the basic solution solver, the reduced basis expansion method and transforming the parameterized inputs into parameterized coefficients of the Nonlinear Schrödinger Equations, universal solutions with respect to inputs corresponding to different bit rates can all be obtained without the need of re-training the whole model. Once adopted, this model can have prominent advantages in both computation efficiency and physical background. Besides, this model can still be effectively trained without the need to collect transmitted signals in advance. Tasks of on-off keying and pulse amplitude modulated signals with various symbol rates over 100 km are utilized to demonstrate the fidelity of the model.