A GNN with on-the-fly random diagonal weights and a frozen pretrained embedding matches end-to-end trained GNNs on several graph benchmarks while reducing training time and memory.
A unified lottery ticket hypothesis for graph neural networks
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
On the Effectiveness of Random Weights in Graph Neural Networks
A GNN with on-the-fly random diagonal weights and a frozen pretrained embedding matches end-to-end trained GNNs on several graph benchmarks while reducing training time and memory.