Skip to content

This repository contains the official implementation code of the paper Improving Multimodal Fusion with Hierarchical Mutual Information Maximization for Multimodal Sentiment Analysis, accepted at EMNLP 2021.

License

Notifications You must be signed in to change notification settings

declare-lab/Multimodal-Infomax

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

12 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

MultiModal-InfoMax

This repository contains the official implementation code of the paper Improving Multimodal Fusion with Hierarchical Mutual Information Maximization for Multimodal Sentiment Analysis, accepted at EMNLP 2021.

🔥 If you would be interested in other multimodal works in our DeCLaRe Lab, welcome to visit the clustered repository

Introduction

Multimodal-informax (MMIM) synthesizes fusion results from multi-modality input through a two-level mutual information (MI) maximization. We use BA (Barber-Agakov) lower bound and contrastive predictive coding as the target function to be maximized. To facilitate the computation, we design an entropy estimation module with associated history data memory to facilitate the computation of BA lower bound and the training process.

Alt text

Usage

  1. Download the CMU-MOSI and CMU-MOSEI dataset from Google Drive or Baidu Disk (extraction code: g3m2). Place them under the folder Multimodal-Infomax/datasets

  2. Set up the environment (need conda prerequisite)

conda env create -f environment.yml
conda activate MMIM
  1. Start training
python main.py --dataset mosi --contrast

Citation

Please cite our paper if you find our work useful for your research:

@inproceedings{han2021improving,
  title={Improving Multimodal Fusion with Hierarchical Mutual Information Maximization for Multimodal Sentiment Analysis},
  author={Han, Wei and Chen, Hui and Poria, Soujanya},
  booktitle={Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing},
  pages={9180--9192},
  year={2021}
}

Contact

Should you have any question, feel free to contact me through henryhan88888@gmail.com

About

This repository contains the official implementation code of the paper Improving Multimodal Fusion with Hierarchical Mutual Information Maximization for Multimodal Sentiment Analysis, accepted at EMNLP 2021.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages