Hello,
Actually, I'm a social scientist. Regarding this, I'm trying to deploy your work to analyse and forecast social interaction.
Therefore, the number of nodes is tremendously large in my dataset. Since you wrote on Readme.MD that you don't deploy the network when the number of edges is relatively smaller than the number of nodes. I wonder is there any desirable ratio between the number of nodes and edges to yield high accuracy.
Plus, when I enlarge the number of nodes, should training time increase linearly or exponentially? I'm curious about this models time complexity in big-O notation please share if you know.
Thank you for sharing the powerful Graph Network model and I will definitely cite your work when I submit my own research deploying your idea.
Hello, Actually, I'm a social scientist. Regarding this, I'm trying to deploy your work to analyse and forecast social interaction. Therefore, the number of nodes is tremendously large in my dataset. Since you wrote on Readme.MD that you don't deploy the network when the number of edges is relatively smaller than the number of nodes. I wonder is there any desirable ratio between the number of nodes and edges to yield high accuracy. Plus, when I enlarge the number of nodes, should training time increase linearly or exponentially? I'm curious about this models time complexity in big-O notation please share if you know. Thank you for sharing the powerful Graph Network model and I will definitely cite your work when I submit my own research deploying your idea.