Robust Spatial Filtering With Graph Convolutional Neural Networks

By Felipe Petroski Such, Shagan Sah, Miguel Domínguez, Suhas Pillai, Andrew M. Michael, Nathan D. Cahill, R. Ptucha. 2017

142 citations recorded by the source corpus.

Convolutional neural networks (CNNs) have recently led to incredible breakthroughs on a variety of pattern recognition problems. Banks of finite-impulse response filters are learned on a hierarchy of layers, each contributing more abstract information than the previous layer. The simplicity and elegance of the convolutional filtering process makes them perfect for structured problems, such as image, video, or voice, where vertices are homogeneous in the sense of number, location, and strength of neighbors. The vast majority of classification problems, for example in the pharmaceutical, homeland security, and financial domains are unstructured. As these problems are formulated into unstructured graphs, the heterogeneity of these problems, such as number of vertices, number of connections per vertex, and edge strength, cannot be tackled with standard convolutional techniques. We propose a novel neural learning framework that is capable of handling both homogeneous and heterogeneous data while retaining the benefits of traditional CNN successes. Recently, researchers have proposed variations of CNNs that can handle graph data. In an effort to create learnable filter banks of graphs, these methods either induce constraints on the data or require preprocessing. As opposed to spectral methods, our framework, which we term Graph-CNNs, defines filters as polynomials of functions of the graph adjacency matrix. Graph-CNNs can handle both heterogeneous and homogeneous graph data, including graphs having entirely different vertex or edge sets. We perform experiments to validate the applicability of Graph-CNNs to a variety of structured and unstructured classification problems and demonstrate state-of-the-art results on document and molecule classification problems.

Subfields: Artificial Intelligence

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

Attention-based Graph Neural Network for Semi-supervised Learning (2018, methodology): “3-fold Split 10-fold split Graph-CNN [38] 87.55 ± 1 .”
Adaptive Sampling Towards Fast Graph Representation Learning (2018, methodology): “To allow the network to better utilize information across distant nodes, we can sample the multi-hop neighborhoods for the GCN update in a similar way as the random walk [6, 10].”
Spectral–Spatial Graph Convolutional Networks for Semisupervised Hyperspectral Image Classification (2019, methodology): “However, directly applying the CNN to a graph to produce useful embedding or feature is a great challenge [9].”
Adaptively Connected Neural Networks (2019, background): “[38] defined filters as polynomials of functions of the graph adjacency matrix for unstructured graph data.”
EEG Based Emotion Recognition: A Tutorial and Review (2022, methodology): “The graph-based description method is advantageous in extracting signals’ discriminative features in the discrete spatial domain [136], the structural representations learned from the functional connectivity graph could capture those correlation information mentioned above.”
MeshCNN: a network with an edge (2019, background): “Such et al. [2017] introduced the concept of vertex filtering on graphs, but did not incorporate pooling operations for feature aggregation.”
Spatial Temporal Transformer Network for Skeleton-based Action Recognition (2020, background): “Another approach is the spatial one, where graph convolution is defined as information aggregation [37]– [39].”
Skeleton-based action recognition via spatial and temporal transformer networks (2020, background): “A complementary approach is the spatial one, where graph convolution is defined as information aggregation [33, 35, 46].”
EEG Emotion Recognition Using Dynamical Graph Convolutional Neural Networks (2020, methodology): “Compared with classical CNN method, GCNN would be more advantageous in dealing with the discriminative feature extraction of signals in the discrete spatial domain [46].”
Classifying Signals on Irregular Domains via Convolutional Cluster Pooling (2019, background): “and Graph-CNNs [25] - to the best of our knowledge, the only graph classification model in literature that reports results on CIFAR-10 - by a significant margin.”
General-Purpose Deep Point Cloud Feature Extractor (2018, methodology): “Some researchers create a graph from the data so that the points benefit from connections that encode structure, either by treating the mesh as a graph [49, 18], or drawing a k-nearest neighbor graph from the point clouds [45, 35].”