Integrating Parallel Computation Technique to State-Space Dynamic Neural Network Modeling of Optical Fiber

In this paper, the previously known state-space dynamic neural network has been formulated such that it can be suitable for parallel computation. Using the parallelization method makes the model training extremely faster. In addition, the final models created using this technique have much smaller evaluation times compared to conventional models present in traditional circuit simulation tools. Finally, the proposed method has been validated through transient modeling of 100KM optical fiber.