mmgcl's Introduction

MMGCL

Source code of our MMGCL

Multi-modal Graph Contrastive Learning for Micro-video Recommendation

Citation

If you want to use our codes in your research, please cite:

@inproceedings{yi2022multi,
  title={Multi-modal graph contrastive learning for micro-video recommendation},
  author={Yi, Zixuan and Wang, Xi and Ounis, Iadh and Macdonald, Craig},
  booktitle={Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval},
  pages={1807--1811},
  year={2022}
}

Environment

The codes of PGCL are implemented and tested under the following development environment:

numba==0.53.1
numpy==1.20.3
scipy==1.6.2
torch==1.10.3
torch_scatter==2.0.6

Usage

Configure the xx.conf file in the directory named conf. (xx is the name of the model you want to run)
Run main.py and choose the model you want to run.

Datasets

You can find the full version of recommendation datasets via Tiktok, and Movielens. Since the copyright of datasets, we cannot release them directly.

Update:

We tend to submit this version to the multi modal recommendation library shortly.

mmgcl's People

Contributors

Stargazers

Watchers

mmgcl's Issues

关于对比损失上的问题

我看到代码中的对比损失是以下内容进行对比：
logits_1 = torch.mm(users_sub_1, items_sub_1.T)
logits_1 /= self.ssl_temp
labels_1 = torch.tensor(list(range(users_sub_2.shape[0]))).to(self.device)
ssl_loss_1 = F.cross_entropy(logits_1, labels_1)

        logits_2 = torch.mm(users_sub_1, items_sub_1.T)
        logits_2 /= self.ssl_temp
        labels_2 = torch.tensor(list(range(users_sub_2.shape[0]))).to(self.device)
        ssl_loss_2 = F.cross_entropy(logits_2, labels_2)

        logits_3 = torch.mm(users_sub_1, neg_items_sub_1.T)
        logits_3 /= self.ssl_temp
        labels_3 = torch.tensor(list(range(users_sub_2.shape[0]))).to(self.device)
        ssl_loss_3 = - F.cross_entropy(logits_3, labels_3)
        ssl_loss = ssl_loss_1 + ssl_loss_2

首先ssl_loss_1 和ssl_loss_2 计算出来的难道不重复吗，其次我看论文中的是用户侧一个损失，项目测一个损失，是否是以下的代码：
logits_1 = torch.mm(users_sub_1, users_sub_2.T) / self.ssl_temp
labels_1 = torch.tensor(list(range(users_sub_1.shape[0]))).to(self.device)
ssl_loss_1 = F.cross_entropy(logits_1, labels_1)

        logits_2 = torch.mm(items_sub_1, items_sub_2.T) / self.ssl_temp
        labels_2 = torch.tensor(list(range(items_sub_1.shape[0]))).to(self.device)
        ssl_loss_2 = F.cross_entropy(logits_2, labels_2)
        ssl_loss = ssl_loss_1 + ssl_loss_2

Recommend Projects

zxy-ml84 / mmgcl Goto Github PK

mmgcl's Introduction

MMGCL

Multi-modal Graph Contrastive Learning for Micro-video Recommendation

Citation

Environment

Usage

Datasets

Update:

mmgcl's People

Contributors

Stargazers

Watchers

Forkers

mmgcl's Issues

关于对比损失上的问题

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent