Dmitry Korobchenko - PyTorch Geometric for Graph Neural Nets | PyData Yerevan 2022

[@theodorosathanasiadis7489]

0:00 Welcome 0:09 Introduction and Agenda 1:36 Deep Learning Basics 2:38 Data and Models 3:51 Shortcomings of MLPs, CNNs, and RNNs for Graphs 5:10 Examples and tasks with Graphs 6:15 Terminology of graphs - Vertices and Edges 7:30 Adjacency matrix for graph conectivity 8:49 Coordinate format for graph conectivity 9:34 Graph Neural Networks 10:26 Node feature vectors 11:29 Processing graphs 13:04 Message passing 14:19 Message function 14:55 Aggregation of message functions 15:36 Update function 16:45 Message passing example 18:52 GNN Variations 22:17 PyTorch Geometric (PyG) Introduction 23:19 PyG Enzymes Dataset 24:53 PyG Data Loader 27:26 PyG GNN Model 28:08 GNN Model constructor method 29:11 GNN forward method 31:37 GNN model instatiation 31:48 PyG Training Loop 32:20 PyG Custom Message Passing 35:03 QnA Session