Graph Convolutional Network with Bootstrap Your Own Latent for Modulation Format Classification in Optical Networks

Nowadays, the Optical Networks (ONs) are undergoing quick evolution to deliver innovative solutions capable of handling the explosive growth in data traffic demands. The rapid growth in Information Technology (IT) and soaring data demands have introduced new complications in ONs, particularly in bandwidth capacity, latency and overall network reliability. However, the existing model based on Singular Value decomposition (SVM) with Radon Transform (RT) for Modulated Formats (MF) classification, but it has required high computational complexity. Hence, this research proposes a Graph Convolutional Network (GCN) with Bootstrap Your Own Latent (BYOL) for the classification of MF in ONs. Initially the data is collected from RadioML2018.01A dataset of modulations and preprocessed with two techniques namely Infinite Impulse Response (IIR) and Min-Max scaling. Then, the features are extracted with the help of BYOL, as it extracts the features based on similarity. Finally, GCN classifier is introduced to classify the MF in ONs, where it provides insights into relationships between various parts of signals. As per the results, the proposed GCN-BYOL model provided outperforming results in accuracy with 98.7% when compared with existing Modulation Format Identification and Classification in Super-Channel Optical Networks (MFIC-SCON).