Using Graph Neural Networks to improve flavour-tagging and its modeling for the measurement of $WH$ and $ZH$ production in the $H\rightarrow bb$ and $H\rightarrow cc$ decay channel.
Graph Neural Networks (GNNs) are machine learning algorithms particularly suitable for modeling data with complex topological correlations. In this communication, their application in the measurement of $WH$ and $ZH$ production in the $H\rightarrow bb$ and $H\rightarrow cc$ decay channel is presented. First, a new GNN algorithm which is being developed in ATLAS to identify jets containing $b$-hadrons by representing them as graphs of tracks and silicon hits is illustrated. Then, a GNN-based technique for weighting events according to the probability of the jets being $b$-tagged is presented.