Skip to content
← All publications

Generalizing CNNs to graphs with learnable neighborhood quantization

I. Osafo Nkansah, N. Gallagher, R. Sandilya, C. Liston, L. Grosenick (2024).

Advances in Neural Information Processing Systems (NeurIPS)

Paper Code PDF ★ Top-3 AI/ML venue

In brief

Quantized Graph Convolution Networks (QGCNs) extend convolutional neural networks directly to graph-structured data, such as brain networks, while reducing exactly to a standard CNN layer on image-like data.

Abstract

Convolutional neural networks (CNNs) have led to a revolution in analyzing array data. However, many important sources of data, such as biological and social networks, are naturally structured as graphs rather than arrays, making the design of graph neural network (GNN) architectures that retain the strengths of CNNs an active and exciting area of research. Here, we introduce Quantized Graph Convolution Networks (QGCNs), the first framework for GNNs that formally and directly extends CNNs to graphs. QGCNs do this by decomposing the convolution operation into non-overlapping sub-kernels, allowing them to fit graph data while reducing to a 2D CNN layer on array data. We generalize this approach to graphs of arbitrary size and dimension by approaching sub-kernel assignment as a learnable multinomial assignment problem. Integrating this approach into a residual network architecture, we demonstrate performance that matches or exceeds other state-of-the-art GNNs on benchmark graph datasets and for predicting properties of nonlinear dynamics on a new finite element graph dataset. In summary, QGCNs are a novel GNN framework that generalizes CNNs and their strengths to graph data, allowing for more accurate and expressive models.

Cite

I. Osafo Nkansah, N. Gallagher, R. Sandilya, C. Liston, L. Grosenick (2024). Generalizing CNNs to graphs with learnable neighborhood quantization. Advances in Neural Information Processing Systems (NeurIPS).