Self-Attention Mechanism-Based Multi-Channel QoT Estimation in Optical Networks

It is essential to estimate the quality of transmission (QoT) lightpaths before their establishment for efficient planning and operation optical networks. Due nonlinear effect fibers, deployed influence QoT each other; thus, multi-channel estimation necessary, which provides complete information network optimization. Moreover, different interfering channels have effects on channel under test. However, existing artificial-neural-network-based estimators (ANN-QoT-E) neglect in input layer, affects accuracy severely. In this paper, we propose a self-attention mechanism-based estimator (SA-QoT-E) improve ANN-QoT-E. SA-QoT-E, features are designed as sequence feature vectors that route same path, mechanism dynamically assigns weights according hyperparameter search method used optimize SA-QoT-E. The simulation results show that, compared with ANN-QoT-E, our proposed SA-QoT-E achieves higher accuracy, can be directly applied wavelength expansion scenarios without retraining.