Bayesian Semi-supervised Learning with Graph Gaussian Processes
This addresses data-efficient learning for graph-based tasks, offering a robust alternative to neural networks in label-scarce scenarios, though it appears incremental as it builds on existing Gaussian process and graph methods.
The authors tackled the semi-supervised learning problem on graphs by proposing a Bayesian Gaussian process-based model, which achieved competitive performance against state-of-the-art graph neural networks on benchmarks and outperformed them in active learning with scarce labels, without needing validation data for early stopping.
We propose a data-efficient Gaussian process-based Bayesian approach to the semi-supervised learning problem on graphs. The proposed model shows extremely competitive performance when compared to the state-of-the-art graph neural networks on semi-supervised learning benchmark experiments, and outperforms the neural networks in active learning experiments where labels are scarce. Furthermore, the model does not require a validation data set for early stopping to control over-fitting. Our model can be viewed as an instance of empirical distribution regression weighted locally by network connectivity. We further motivate the intuitive construction of the model with a Bayesian linear model interpretation where the node features are filtered by an operator related to the graph Laplacian. The method can be easily implemented by adapting off-the-shelf scalable variational inference algorithms for Gaussian processes.