On the Unreasonable Effectiveness of Feature propagation in Learning on Graphs with Missing Node Features
This addresses a practical bottleneck for applying GNNs in domains with incomplete data, offering a scalable solution with strong empirical results.
The paper tackles the problem of missing node features in graph neural networks, which is common in real-world applications like social networks, by proposing Feature Propagation, a simple diffusion-based algorithm that minimizes Dirichlet energy. The method outperforms previous approaches on seven node-classification benchmarks, showing only about 4% relative accuracy drop even with 99% missing features and running in 10 seconds on a large graph with ~2.5M nodes.
While Graph Neural Networks (GNNs) have recently become the de facto standard for modeling relational data, they impose a strong assumption on the availability of the node or edge features of the graph. In many real-world applications, however, features are only partially available; for example, in social networks, age and gender are available only for a small subset of users. We present a general approach for handling missing features in graph machine learning applications that is based on minimization of the Dirichlet energy and leads to a diffusion-type differential equation on the graph. The discretization of this equation produces a simple, fast and scalable algorithm which we call Feature Propagation. We experimentally show that the proposed approach outperforms previous methods on seven common node-classification benchmarks and can withstand surprisingly high rates of missing features: on average we observe only around 4% relative accuracy drop when 99% of the features are missing. Moreover, it takes only 10 seconds to run on a graph with $\sim$2.5M nodes and $\sim$123M edges on a single GPU.