Pojdi na vsebino
- Learning Tasks on Graphs
- Basics of deep learning on graphs
- Node-, link- & (sub)graph-level tasks
- Transductive & inductive learning setting
- Real-world examples & applications
- Graph Prediction without Learning
- Node ranking by position & centrality
- Node clustering by community detection
- Link prediction by node similarity indices
- Graph similarity by kernel functions
- Graph Representation Learning
- “Shallow” encoder & decoder
- Matrix factorization methods
- Random-walk (node) embeddings
- DeepWalk, LINE, node2vec, etc.
- Graph Neural Networks (GNN)
- Basic architecture (message, aggregation, layers)
- Permutation invariance & equivariance
- Over-smoothing & -squashing problems
- GCN, GraphSAGE, GAT, GIN, etc.
- GNN Training in Practice
- Graph splitting (train, validation & test sets)
- GNN design space & graph augmentation
- Loss, training, prediction & evaluation
- GNN Theory & Limitations
- Weisfeiler-Leman graph kernel
- Structure- & position-awareness
- Introduction to Graph Transformers
- Special Graphs & Applications
- Heterogeneous graphs (RGCN, HGT, etc.)
- Recommender systems (NGCF, LightGCN, etc.)
- Knowledge graphs (TransE, RotateE, etc.)
- Selected other applications
- Relational Deep Learning
- Distribution of hours per semester
Instructor
Room:R2.49 - Kabinet
