Please enable JavaScript.
Coggle requires JavaScript to display documents.
GNN, other graph types, different training settings - Coggle Diagram
GNN
-
3) Design loss function
-
Graph-level (graph classification, graph regression, and graph matching)
Node-level (node classification, node regression, node clustering)
-
-
-
-
-
other graph types
Dynamics graphes
graph structure, e.g. the existence of edges and nodes, keeps changing over time.
Large graph :
Leveraging approximate personalized PageRank, methods proposed by Klicpera et al. (2019) and Bojchevski et al. (2020) avoid calculating high-order propagation matrices. Rossi et al. (2020) propose a method to precompute graph convolutional filters of different sizes for efficient training and inference.
signed graphs : graphs with signed edges, i.e. an edge can be either positive or negative)
-