Gradient-based Optimization for Unrepeatered Optical Systems

Abstract

We propose a multi-parameter adaptive technique based on the gradient descent algorithm (MAT-GRAD) for optimizing the amplification scheme in unrepeatered wavelength-division multiplexed optical systems. The proposed methodology uses the gradient descent method to jointly optimize the transmitter (Tx) and receiver (Rx) remote optically pumped amplifiers (ROPAs), as well as the Tx forward first-order distributed Raman amplifier (DRA). The MAT-GRAD algorithm is evaluated in combination with the conventional gradient descent algorithm (GDA), and GDA with momentum (GDAM). The simulations indicate that the MAT-GRAD with GDAM achieves a suitable minimum at a higher convergence speed.