Supervised Graph Neural Network - Week 7
The goal is to understand how GNNs work.
Created on November 10|Last edited on November 29
Comment
The environment where data is collected.
Corresponding graph.
Task: Given each node has the absolute position, predict the relative position.
Overfitting to one batch: Max Pooling
Overfitting to one batch: Node Aggregation
Overfitting Insufficient to Test Message Passing
Training on FullDataset: Edge Type
Bi-Directional connection Exploration
Residual Networks
Wanted to compare how residual networks function. With, without self loops and different edge-types.
Single Layer GNN
Number of Layers
Number of layers. There doesn't seem to be a pattern I could understand. And it is somewhat probabilistic whether the networks learns properly.
XXX OPEN QUESTIONS XXX
Answers
Add a comment