Skip to content
/ PMMM Public

Differentiable Meta Multigraph Search with Partial Message Propagation on Heterogeneous Information Networks

License

Notifications You must be signed in to change notification settings

JHL-HUST/PMMM

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

37 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

PMMM

We provide code for the Paper:

Differentiable Meta Multigraph Search with Partial Message Propagation on Heterogeneous Information Networks
Chao Li, Hao Xu, Kun He
[AAAI 2023]

Introduction

Our main contributions are as follows:

  1. To our knowledge, PMMM is the first NAS method to search multigraph as the architecture. We propose a new concept of meta multigraph to guide GNNs to propagate messages for NAS on HINs.

  2. We propose the first stable differentiable architecture search algorithm on HINs, called partial message search, which can consistently discover promising architectures that outperform hand-designed models.

  3. Thorough experiments are conducted to demonstrate the effectiveness, stability, and flexibility of our method.

Requisites

python=3.8
pytorch==1.6.0 with CUDA support (by default our model is trained on GPU)
numpy==1.19.1
scipy==1.5.2
pandas==1.1.1
scikit-learn==0.23.2

Quick Start

For the node classification task, please see README under node_classification, and for the recommendation task, please see README under recommendation.

Citation

If you find this code and data useful, please consider citing the original work by authors:

@article{li2022differentiable,
  title={Differentiable Meta Multigraph Search with Partial Message Propagation on Heterogeneous Information Networks},
  author={Li, Chao and Xu, Hao and He, Kun},
  journal={arXiv preprint arXiv:2211.14752},
  year={2022}
}

About

Differentiable Meta Multigraph Search with Partial Message Propagation on Heterogeneous Information Networks

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 3

  •  
  •  
  •