MGAT

This is our Pytorch implementation for our paper- Multimodal Graph Attention Network(MGAT):

Zhulin Tao, Yinwei Wei, Xiang Wang, Xiangnan He, Xianglin Huang, Tat-Seng Chua: MGAT: Multimodal Graph Attention Network for Recommendation. Inf. Process. Manag. 57(5): 102277 (2020)

Introduction

In this work, we propose a new Multimodal Graph Attention Network, short for MGAT, which disentangles personal interests at the granularity of modality. In particular, built upon multimodal interaction graphs, MGAT conducts information propagation within individual graphs, while leveraging the gated attention mechanism to identify varying importance scores of different modalities to user preference.

Environment Requirement

The code has been tested running under Python 3.6.5. The required packages are as follows:

• torch==1.7.0
• numpy==1.16.1
• torch_geometric==1.6.1

run

CUDA_VISIBLE_DEVICES=0 python  -u train.py --num_epoch 200 --batch_size 2048 --weight_decay 0.1 --l_r 3e-5

Citation

@article{DBLP:journals/ipm/TaoWWHHC20, author = {Zhulin Tao and Yinwei Wei and Xiang Wang and Xiangnan He and Xianglin Huang and Tat{-}Seng Chua}, title = {{MGAT:} Multimodal Graph Attention Network for Recommendation}, journal = {Inf. Process. Manag.}, volume = {57}, number = {5}, pages = {102277}, year = {2020} }