BM3 (WWW'23)

Pytorch implementation for "Bootstrap Latent Representations for Multi-modal Recommendation"-WWW'23 Official ACM

• Trained logs&models are stored at: https://github.com/enoche/BM3/tree/master/trained-models-logs
• 🔀 This model is integrated into the MMRec framework.
• 👉 Check the awesome multimodal recommendation resources.

Download from Google Drive: Baby/Sports/Elec
The data already contains text and image features extracted from Sentence-Transformers and CNN.

1. Put your downloaded data (e.g. baby) under data dir.
2. Enter src folder and run with
   python main.py -m BM3 -d baby
   You may specify other parameters in CMD or config with configs/model/*.yaml and configs/dataset/*.yaml.

We report the best hyper-parameters of BM3 to reproduce the results in Table III of our paper as:

@inproceedings{zhou2023bootstrap,
author = {Zhou, Xin and Zhou, Hongyu and Liu, Yong and Zeng, Zhiwei and Miao, Chunyan and Wang, Pengwei and You, Yuan and Jiang, Feijun},
title = {Bootstrap Latent Representations for Multi-Modal Recommendation},
booktitle = {Proceedings of the ACM Web Conference 2023},
pages = {845–854},
year = {2023}
}