Robustness to Missing Features using Hierarchical Clustering with Split Neural Networks

Code for the paper - https://ojs.aaai.org/index.php/AAAI/article/view/17905

The following readme has simple steps to reproduce the training and evaluation for any of the datasets mentioned.

Setup

1. Setup Virtual Environment

pip install virtualenv
virtualenv venv
source venv/bin/activate

2. Install dependencies pip install -r requirements.txt

3. Run the code

Run

Train

python main.py train --dataset_dir datasets --dataset life --model_dir life_models --verbose 1

Evaluate

python main.py evaluate --dataset_dir datasets --dataset life --model_dir life_models --verbose 1

Further Notes

Mapping for datasets to --dataset flag

1. Life Expectancy (WHO) : life
2. Bands : bands
3. Kidney Disease : kidney_disease
4. Mammographics : mammographics
5. Horse Colic : horse
6. Pima Indians : pima
7. Hepatitis : hepatitis
8. Breast Cancer Winconsin : winconsin

Citation

If you find this project useful for your research, please use the following BibTeX entry to cite our paper https://ojs.aaai.org/index.php/AAAI/article/view/17905.

@article{khincha2021missing,
        author={Khincha, Rishab and Sarawgi, Utkarsh and Zulfikar, Wazeer and Maes, Pattie}, 
        title={Robustness to Missing Features using Hierarchical Clustering with Split Neural Networks (Student Abstract)}, 
        volume={35}, 
        url={https://ojs.aaai.org/index.php/AAAI/article/view/17905}, 
        number={18}, 
        journal={Proceedings of the AAAI Conference on Artificial Intelligence}, 
        year={2021}, 
        month={May}, 
        pages={15817-15818}
}