Haar Wavelet Feature Compression for Quantized Graph Convolutional Networks

TL;DR

This paper introduces Haar wavelet-based compression combined with light quantization to reduce computational costs in Graph Convolutional Networks, outperforming traditional aggressive quantization across various graph-based tasks.

Contribution

The paper proposes a novel Haar wavelet compression method for GCN features, improving efficiency while maintaining performance compared to aggressive quantization.

Findings

Haar wavelet compression significantly reduces computation and bandwidth.

The approach outperforms aggressive quantization in multiple tasks.

Effective for node classification, point cloud classification, and segmentation.

Abstract

Graph Convolutional Networks (GCNs) are widely used in a variety of applications, and can be seen as an unstructured version of standard Convolutional Neural Networks (CNNs). As in CNNs, the computational cost of GCNs for large input graphs (such as large point clouds or meshes) can be high and inhibit the use of these networks, especially in environments with low computational resources. To ease these costs, quantization can be applied to GCNs. However, aggressive quantization of the feature maps can lead to a significant degradation in performance. On a different note, Haar wavelet transforms are known to be one of the most effective and efficient approaches to compress signals. Therefore, instead of applying aggressive quantization to feature maps, we propose to utilize Haar wavelet compression and light quantization to reduce the computations and the bandwidth involved with the network. We demonstrate that this approach surpasses aggressive feature quantization by a significant margin, for a variety of problems ranging from node classification to point cloud classification and part and semantic segmentation.