Unveiling-the-potential-of-GNN-for-network-modeling-and-optimization-in-SDN

Unveiling-the-potential-of-GNN-for-network-modeling-and-optimization-in-SDN copied to clipboard

Hello, I have used the trained model to predict delay and jitter, the results are satisfying. But when I use RoueNet-SOSR.py with target "drops" to train and use the new model to predict drops, the result is so bad... The datasets and training steps that I used are all totally from your Readme.md, I have tested the model on nsfnet. So is there anything wrong during my experiment ? Or what should I do if I want to have a more precise prediction of drops? Thanks.