Prioritization method of link protection with convolutional neural networks

In order to provide high quality services to users, communication networks need to keep high reliability. Therefore, they are operated so as to have tolerance to node and link failures. In this paper, we propose a prioritization method of link protection using convolutional neural networks, which are applied to a lot of research fields owing to its dramatic performance. To decide which links should be protected, we calculate the network availability. However, in general, the calculation time of the network availability exponentially increases as the number of links increases. The proposed method gives the priority of link to be protected in a short time by convolutional neural networks, where weighted adjacency matrices of nodes are used as input data. Through numerical experiments with CNN, we show the fundamental efficiency of our prioritization method.