Data

Tools

Indexes

Applications

Experiments

Open Source Science

unifews

The original code for ICML 2025 paper "Unifews: You Need Fewer Operations for Efficient Graph Neural Networks"

https://github.com/gdmnl/unifews

Science Score: 54.0%

This score indicates how likely this project is to be science-related based on various indicators:

  • ✓
    CITATION.cff file
    Found CITATION.cff file
  • ✓
    codemeta.json file
    Found codemeta.json file
  • ✓
    .zenodo.json file
    Found .zenodo.json file
  • ○
    DOI references
  • ✓
    Academic publication links
    Links to: arxiv.org
  • ○
    Academic email domains
  • ○
    Institutional organization owner
  • ○
    JOSS paper metadata
  • ○
    Scientific vocabulary similarity
    Low similarity (11.3%) to scientific vocabulary
Last synced: 10 months ago · JSON representation ·

Repository

The original code for ICML 2025 paper "Unifews: You Need Fewer Operations for Efficient Graph Neural Networks"

Basic Info
  • Host: GitHub
  • Owner: gdmnl
  • Language: Python
  • Default Branch: main
  • Homepage:
  • Size: 1.8 MB
Statistics
  • Stars: 2
  • Watchers: 1
  • Forks: 0
  • Open Issues: 0
  • Releases: 0
Created over 2 years ago · Last pushed 11 months ago
Metadata Files
Readme Citation

README.md

Unifews

This is the original code for Unifews: Unified Entry-Wise Sparsification for Efficient Graph Neural Network, ICML 2025.

Conference (Poster/Video/Slides) | OpenReview | arXiv | GitHub

Citation

If you find this work useful, please cite our paper:

Ningyi Liao, Zihao Yu, Ruixiao Zeng, and Siqiang Luo.
Unifews: You Need Fewer Operations for Efficient Graph Neural Networks.
In Proceedings of the 42nd International Conference on Machine Learning, PMLR 267, 2025. @inproceedings{liao2025unifews, title={{Unifews}: You Need Fewer Operations for Efficient Graph Neural Networks}, author={Liao, Ningyi and Yu, Zihao and Ruixiao Zeng and Luo, Siqiang}, booktitle={42nd International Conference on Machine Learning}, year={2025}, month={May}, publisher={PMLR}, volume={267}, location={Vancouver, Canada}, url={https://icml.cc/virtual/2025/poster/45740}, }

Dependencies

Python

Installed env.txt by conda: bash conda create --name <env> --file env.txt

C++

Experiment

Data Preparation

  1. Use utils/data_transfer.py to generate processed files under path data/[dataset_name] similar to the example folder data/cora:
    • adj.npz: adjacency matrix in scipy.sparse.csr_matrix
    • feats.npy: features in .npy array
    • labels.npz: node label information
    • 'label': labels (number or one-hot)
    • 'idxtrain/idxval/idx_test': indices of training/validation/test nodes
    • adj_el.bin, adj_pl.bin, attribute.txt, degree.npz: graph files for precomputation

Decoupled Model Propagation

  1. Compile Cython: bash cd precompute python setup.py build_ext --inplace

Model Training

  1. Run full-batch experiment: bash python run_fb.py -f [seed] -c [config_file] -v [device]
  2. Run mini-batch experiment bash python run_mb.py -f [seed] -c [config_file] -v [device]

Reference & Links

Datasets

Baselines

  • GLT: A Unified Lottery Ticket Hypothesis for Graph Neural Networks
  • GEBT: Early-Bird GCNs: Graph-Network Co-optimization towards More Efficient GCN Training and Inference via Drawing Early-Bird Lottery Tickets
  • CGP: Comprehensive Graph Gradual Pruning for Sparse Training in Graph Neural Networks
  • DSpar: DSpar: An Embarrassingly Simple Strategy for Efficient GNN Training and Inference via Degree-Based Sparsification
  • NDLS: Node Dependent Local Smoothing for Scalable Graph Learning
  • NIGCN: Node-wise Diffusion for Scalable Graph Learning

Owner

  • Name: gdmnl
  • Login: gdmnl
  • Kind: organization

Citation (CITATION.cff) 

cff-version: 1.2.0
message: "If you use this framework, please cite our paper as below."
authors:
  - family-names: Liao
    given-names: Ningyi
    orcid: https://orcid.org/0000-0003-3176-4401
  - family-names: Yu
    given-names: Zihao
  - family-names: Zeng
    given-names: Ruixiao
  - family-names: Luo
    given-names: Siqiang
    orcid: https://orcid.org/0000-0001-8197-0903
title: "Unifews"
version: 1.0.0
date-released: 2025-05-30
url: https://icml.cc/virtual/2025/poster/45740
eprint: 2403.13268
preferred-citation:
  type: conference-paper
  authors:
  - family-names: Liao
    given-names: Ningyi
    orcid: https://orcid.org/0000-0003-3176-4401
  - family-names: Yu
    given-names: Zihao
  - family-names: Zeng
    given-names: Ruixiao
  - family-names: Luo
    given-names: Siqiang
    orcid: https://orcid.org/0000-0001-8197-0903
  journal: "42nd International Conference on Machine Learning"
  month: 5
  title: "Unifews: You Need Fewer Operations for Efficient Graph Neural Networks"
  publisher: PMLR
  volume: 267
  year: 2025

GitHub Events

Total
  • Push event: 4
Last Year
  • Push event: 4