A Generalization of Convolutional Neural Networks to Graph-Structured Data
This work addresses the need for effective neural network methods for graph data, which is incremental as it builds on CNNs but adapts them to a new domain.
The paper tackles the problem of extending convolutional neural networks to graph-structured data by proposing a novel spatial convolution based on random walks, analogous to standard CNNs for images. It demonstrates performance on MNIST and challenges state-of-the-art on the Merck molecular activity dataset.
This paper introduces a generalization of Convolutional Neural Networks (CNNs) from low-dimensional grid data, such as images, to graph-structured data. We propose a novel spatial convolution utilizing a random walk to uncover the relations within the input, analogous to the way the standard convolution uses the spatial neighborhood of a pixel on the grid. The convolution has an intuitive interpretation, is efficient and scalable and can also be used on data with varying graph structure. Furthermore, this generalization can be applied to many standard regression or classification problems, by learning the the underlying graph. We empirically demonstrate the performance of the proposed CNN on MNIST, and challenge the state-of-the-art on Merck molecular activity data set.