ResMultiHGNN

This repository contains the source code for the paper Residual Enhanced Multi-Hypergraph Neural Network, accepted by ICIP 2021.

Citation

If you find this work useful in your research, please consider cite:

@inproceedings{icip21-ResMHGNN,
  title     = {Residual Enhanced Multi-Hypergraph Neural Network},
  author    = {Huang, Jing and Huang, Xiaolin and Yang, Jie},
  booktitle = {International Conference on Image Processing, {ICIP-21}},
  year      = {2021}
}

Getting Started

Prerequisites

Our code requires Python>=3.6.

We recommed using virtual environtment and install the newest versions of Pytorch.

You also need these additional packages:

scipy
numpy
path

Datasets

Please download the precomputed features of ModelNet40 and NTU2012 datasets from HGNN or just clicking the following links.

Extract above files and put them under any directory ($DATA_ROOT) you like.

3D Object Classification Task

We implement the HGNN, MultiHGNN, ResHGNN and ResMultiHGNN. You can change the $model and the layers $layer.

Full Training Lables

python train.py --dataroot=$DATA_ROOT --dataname=ModelNet40  --seed=2  --model-name=$model --nlayer=$layer; 

python train.py --dataroot=$DATA_ROOT --dataname=NTU2012     --seed=1  --model-name=$model --nlayer=$layer;

Balanced Subset of Training Labels

python train.py --dataroot=$DATA_ROOT --dataname=ModelNet40  --model-name=$model --nlayer=$layer --balanced; 

python train.py --dataroot=$DATA_ROOT --dataname=NTU2012     --model-name=$model --nlayer=$layer --balanced;

Stability Analysis

Change the split-ratio as you like.

python train.py --dataroot=$DATA_ROOT --dataname=ModelNet40   --model-name=$model --nlayer=$layer --split-ratio=4; 

python train.py --dataroot=$DATA_ROOT --dataname=NTU2012      --model-name=$model --nlayer=$layer --split-ratio=4;

Usage

usage: ResMultiHGNN [-h] [--dataroot DATAROOT] [--dataname DATANAME]
                    [--model-name MODEL_NAME] [--nlayer NLAYER] [--nhid NHID]
                    [--dropout DROPOUT] [--epochs EPOCHS]
                    [--patience PATIENCE] [--gpu GPU] [--seed SEED]
                    [--nostdout] [--balanced] [--split-ratio SPLIT_RATIO]
                    [--out-dir OUT_DIR]

optional arguments:
  -h, --help            show this help message and exit
  --dataroot DATAROOT   the directary of your .mat data (default:
                        ~/data/HGNN)
  --dataname DATANAME   data name (ModelNet40/NTU2012) (default: NTU2012)
  --model-name MODEL_NAME
                        (HGNN, ResHGNN, MultiHGNN, ResMultiHGNN) (default:
                        HGNN)
  --nlayer NLAYER       number of hidden layers (default: 2)
  --nhid NHID           number of hidden features (default: 128)
  --dropout DROPOUT     dropout probability (default: 0.5)
  --epochs EPOCHS       number of epochs to train (default: 600)
  --patience PATIENCE   early stop after specific epochs (default: 200)
  --gpu GPU             gpu number to use (default: 0)
  --seed SEED           seed for randomness (default: 1)
  --nostdout            do not output logging info to terminal (default:
                        False)
  --balanced            only use the balanced subset of training labels
                        (default: False)
  --split-ratio SPLIT_RATIO
                        if set unzero, this is for Task: Stability Analysis,
                        new total/train ratio (default: 0)

License

Distributed under the MIT License. See LICENSE for more information.

OneForward / ResMHGNN