MIG-GT

Source code and dataset of the paper "Modality-Independent Graph Neural Networks with Global Transformers for Multimodal Recommendation", which is accepted by AAAI 2025.

Homepage and Paper

• Homepage (MIG-GT): https://github.com/CrawlScript/MIG-GT
• Paper Access:
  • ArXiv: https://arxiv.org/abs/2412.13994

Dataset

The dataset is the same as the one used in the paper 'A Tale of Two Graphs: Freezing and Denoising Graph Structures for Multimodal Recommendation.' Please refer to their official repository to download the pre-processed dataset, which should be placed in the data directory.

Requirements

• Linux
• Python 3.7
• torch==1.12.1+cu113
• torchmetrics==0.11.4
• dgl==1.0.2+cu113
• ogb==1.3.5
• shortuuid==1.0.11
• pandas==1.3.5
• numpy==1.21.6
• tqdm==4.64.1

RUN

# Run the following command:
python main.py --gpu 0 --seed 1 --dataset $DATASET --result_dir results --method mig_gt
# Note: $DATASET can be 'baby', 'sports', or 'clothing'.