/CoCN

Official implementation of Compressed Convolution Networks (CoCNs), including the source code for the ICML 2023 paper "All in a Row: Compressed Convolution Networks for Graphs".

Primary LanguagePython

Compressed Convolution Networks for Graphs (CoCN)

This is the official implementation for Compressed Convolution Networks (CoCN), a general backbone for graph representation learning.

CoCN Highlight

Getting Started

  • CoCN is implemented with permutation generation and diagonal convolution.
  • For the Vanilla CoCN, please refer to Vanilla Ver..
  • We also seek to extend CoCN to encompass a broader range of graph tasks, please refer to General Ver. for our extension. General CoCN enhances the vanilla version in several aspects:
    • Enabling support for graphs with implicit node features.
    • Enhancing scalability through sparsification and segmentation techniques.