ProtoMix （KDD 2024）

This is a Pytorch implementation of ProtoMix: Augmenting Health Status Representation Learning via Prototype-based Mixup. More details of the paper and dataset will be released after it is published.

The Code

Requirements

Following the suggested way by PyHealth: https://github.com/sunlabuiuc/PyHealth. Note that pytorch >=1.10.

Folder Structure

└── code-and-data
    ├── models                  # The core source code of our model ProtoMix
    │   |──  RNN.py             # The Backbone of RNN
    │   |──  ProtoMix.py        # The main model of ProtoMix 
    │   |──  model_utils.py     # Utils file
    │   |──  run_utils.py       # Utils file
    ├── parser.py               # Parser Args
    ├── mian_protomix.py        # This is the main file
    └── README.md               # This document

Datasets

Download MIMIC-III & MIMIC-IV & eICU datasets from https://physionet.org/content/;

Citation

@inproceedings{xu_jiang_chu_et_al, title={ProtoMix: Augmenting Health Status Representation Learning via Prototype-based Mixup}, author={Xu, Yongxin and Jiang, Xinke and Chu, Xu and Xiao, Yuzhen and Zhang, Chaohe and Ding, Hongxin and Zhao, Junfeng and Wang, Yasha and Xie, Bing}, booktitle={SIGKDD}, year={2024} }