Optimality of Message-Passing Architectures for Sparse Graphs
This work provides a theoretical foundation for optimal node classification in sparse graphs, which is incremental as it builds on existing statistical models and message-passing architectures.
The authors tackled the node classification problem on sparse, feature-decorated graphs by introducing an asymptotically locally Bayes optimal classifier, which is implementable via a message-passing graph neural network and achieves a generalization error that interpolates between an MLP and a convolution based on graph signal strength.
We study the node classification problem on feature-decorated graphs in the sparse setting, i.e., when the expected degree of a node is $O(1)$ in the number of nodes, in the fixed-dimensional asymptotic regime, i.e., the dimension of the feature data is fixed while the number of nodes is large. Such graphs are typically known to be locally tree-like. We introduce a notion of Bayes optimality for node classification tasks, called asymptotic local Bayes optimality, and compute the optimal classifier according to this criterion for a fairly general statistical data model with arbitrary distributions of the node features and edge connectivity. The optimal classifier is implementable using a message-passing graph neural network architecture. We then compute the generalization error of this classifier and compare its performance against existing learning methods theoretically on a well-studied statistical model with naturally identifiable signal-to-noise ratios (SNRs) in the data. We find that the optimal message-passing architecture interpolates between a standard MLP in the regime of low graph signal and a typical convolution in the regime of high graph signal. Furthermore, we prove a corresponding non-asymptotic result.