logo SBA

ETD

Archivio digitale delle tesi discusse presso l’Università di Pisa

Tesi etd-02062024-173529


Tipo di tesi
Tesi di laurea magistrale
Autore
GALLOPPI, GIULIANO
URN
etd-02062024-173529
Titolo
Validating GNN-based Link Prediction Explanations via Local Subgraphs
Dipartimento
INFORMATICA
Corso di studi
INFORMATICA
Relatori
relatore Prof.ssa Monreale, Anna
relatore Prof. Rossetti, Giulio
relatore Dott. Piaggesi, Simone
Parole chiave
  • Community detection
  • GNNs
  • Link prediction
  • Network analysis
  • XAI
Data inizio appello
23/02/2024
Consultabilità
Non consultabile
Data di rilascio
23/02/2094
Riassunto
This thesis is based on the validation of novel methods for explaining Link Prediction tasks performed by GNNs (Graph Neural Networks), thus falling within the field of XAI and intersecting with graph theory.
Starting from existing state-of-the-art Explainers for GNNs, we use community detection methods for graphs to evaluate the goodness of subgraph-based explanations.
In particular, we evaluate whether community structure is an element that has been used not explicitly, but as a latent variable for the importance of that prediction.
File