Jose Suárez-Varela (DIISM, University of Siena)
May 29, 2019 – 11:00 AM
DIISM, Artificial Intelligence laboratory (room 201), Siena SI
Computer network modeling is a critical component for building future self-driving computer networks, particularly to find optimal configurations that meet the goals set by network administrators. However, existing modeling techniques do not meet the requirements to provide accurate estimations of relevant end-to-end Key Performance Indicators (KPI) such as delay, jitter or packet drops. In this context, Graph Neural Networks (GNN) seem to be a cost-efficient alternative to achieve proper network models. RouteNet, our recent GNN-based model, is able to capture and model the complex relationships between topology, routing and input traffic to produce accurate estimates of the per-source/destination pair mean delay and jitter. GNNs are tailored to learn and model information structured as graphs and as a result, our model is able to generalize over arbitrary topologies, routing schemes and variable traffic intensity. In addition, we present the potential of our GNN model to be combined with Deep Reinforcement Learning algorithms (e.g., Alpha Zero) in order to achieve efficient automatic network operation (e.g., routing optimization).