Elastic Optical Networking enables efficient use of spectral resources at the cost of a large parameter space which needs to be optimized to maximize transmission bandwidth. By formulating this optimization problem as a Markov Decision Process, we show that by using a state of the art Reinforcement Learning algorithm an agent can be trained, which is able to select near optimal parameters for different link conditions within seconds. Furthermore, the trained agent is able to generalize to unseen conditions, removing the need to optimize and train for every possible link scenario.
A support vector machine (SVM) based detection is applied to different equalization schemes for a data center interconnect link using coherent 64 GBd 64-QAM over 100 km standard single mode fiber (SSMF). Without any prior knowledge or heuristic assumptions, the SVM is able to learn and capture the transmission characteristics from only a short training data set. We show that, with the use of suitable kernel functions, the SVM can create nonlinear decision thresholds and reduce the errors caused by nonlinear phase noise (NLPN), laser phase noise, I/Q imbalances and so forth. In order to apply the SVM to 64-QAM we introduce a binary coding SVM, which provides a binary multiclass classification with reduced complexity. We investigate the performance of this SVM and show how it can improve the bit-error rate (BER) of the entire system. After 100 km the fiber-induced nonlinear penalty is reduced by 2 dB at a BER of 3.7 × 10 . Furthermore, we apply a nonlinear Volterra equalizer (NLVE), which is based on the nonlinear Volterra theory, as another method for mitigating nonlinear effects. The combination of SVM and NLVE reduces the large computational complexity of the NLVE and allows more accurate compensation of nonlinear transmission impairments.
We propose a fast machine learning assisted capacity optimization in a point-to-point link of an elastic optical network using a reinforcement learning (RL) algorithm. We demonstrate a RL agent trained on idealized parameters in simulation, and compare the results with classical optimization algorithms such as a genetic algorithm and a simple power sweep. Furthermore, we present the high generalization ability of the RL agent under varying link conditions. Experimental results show that for 8 DWDM channels over nine 50km spans of True Wave Classic NZDSF with a span loss of 12 dB the RL agent is able to cope with these unseen conditions, removing the need to optimize for every possible link scenario during training.