Dynamic Virtual Network Embedding Algorithm Based on Graph Convolution Neural Network and Reinforcement Learning

Network virtualization (NV) is a technology with broad application prospects. Virtual network embedding (VNE) the core orientation of VN, which aims to provide more flexible underlying physical resource allocation for user function requests. The classical VNE problem usually solved by heuristic method, but this method often limits flexibility algorithm and ignores time limit. In addition, partition autonomy domain dynamic characteristics virtual request (VNR) also increase difficulty VNE. This article proposed new type algorithm, applied reinforcement learning (RL) graph neural (GNN) theory especially combination convolutional (GCNN) RL algorithm. Based on self-defined fitness matrix value, we set up objective implementation, realized an efficient effectively reduced degree fragmentation. Finally, used comparison algorithms evaluate method. Simulation experiments verified that based GCNN has good basic characteristics. By changing attributes network, it can be proved flexibility.