Physics-based deep learning for modeling nonlinear pulse propagation in optical fibers

Abstract

A physics-based deep learning (DL) method termed Phynet is proposed for modeling the nonlinear pulse propagation in optical fibers totally independent of the ground truth. The presented Phynet is a combination of a handcrafted neural network and the nonlinear Schrödinger physics model. In particular, Phynet is optimized through physics loss generated by the interaction between the network and the physical model rather than the supervised loss. The inverse pulse propagation problem is leveraged to exemplify the performance of Phynet when in comparison to the typical DL method under the same structure and datasets. The results demonstrate that Phynet is able to precisely restore the initial pulse profiles with varied initial widths and powers, while revealing a similar prediction accuracy compared with the typical DL method. The proposed Phynet method can be expected to break the severe bottleneck of the traditional DL method in terms of relying on abundant labeled data during the training phase, which thus brings new insight for modeling and predicting the nonlinear dynamics of the fibers.

Predicting nonlinear reshaping of periodic signals in optical fibre with a neural network

“…2(a). It is worth noting that recently, many works have explored the application of more advanced ML methods to NLSE-related problems, such as recurrent [5,6,20], convolutional [21][22][23][24], or physics-informed [25][26][27][28] NNs, achieving remarkable performance. In this work, our goal is not to showcase the most powerful ML method for solving the NLSE propagation problem being considered but rather to emphasise that an easily accessible and model-free method [29] can already fit our purpose well.…”

Conditional Recurrent Neural Networks for broad applications in nonlinear optics

“…Nonlinear optical pulse propagation presents another optimal field for harnessing the capabilities of neural networks and machine learning. A number of different architectures and implementations have been proposed in recent years, ranging from physics-informed solutions that rely on knowledge by the network of the governing equations [16], to fully model-free, data-driven methods that can, for instance, predict the outcomes of the numerical solution of the Nonlinear Schrödinger equation [17][18][19], or improve numerical simulations of optical Rogue-waves [20]. Optical pulse propagation in a nonlinear medium depends on the parameters of the input pump as well as the geometry of the medium.…”