Reducing Oversmoothing through Informed Weight Initialization in Graph Neural Networks
This work addresses a key bottleneck in GNNs for researchers and practitioners by enabling deeper networks through reduced oversmoothing, though it is incremental as it builds on existing initialization ideas.
The paper tackles oversmoothing in Graph Neural Networks (GNNs) by proposing a new weight initialization method (G-Init) based on generalizing Kaiming initialization to account for graph topology, which reduces oversmoothing and improves performance in node and graph classification tasks, with experimental validation showing advantages in deep networks and cold-start scenarios.
In this work, we generalize the ideas of Kaiming initialization to Graph Neural Networks (GNNs) and propose a new scheme (G-Init) that reduces oversmoothing, leading to very good results in node and graph classification tasks. GNNs are commonly initialized using methods designed for other types of Neural Networks, overlooking the underlying graph topology. We analyze theoretically the variance of signals flowing forward and gradients flowing backward in the class of convolutional GNNs. We then simplify our analysis to the case of the GCN and propose a new initialization method. Our results indicate that the new method (G-Init) reduces oversmoothing in deep GNNs, facilitating their effective use. Experimental validation supports our theoretical findings, demonstrating the advantages of deep networks in scenarios with no feature information for unlabeled nodes (i.e., ``cold start'' scenario).