Differentiable Learning Under Triage

This is a repository containing code and data for the paper:

N. Okati, A. De, and M. Gomez-Rodriguez. Differentiable Learning Under Triage. Advances in Neural Information Processing Systems (NeurIPS), 2021.

The paper is available here.

Pre-requisites

This code depends on the following packages:

1. jupyterlab
2. Torch
3. fasttext
4. matplotlib
5. numpy

Structure

• Hatespeech contains the data and code for regenerating the experiments on Hatespeech dataset.
• Galaxy-zoo contains the data and code for regenerating the experiments on Galaxy-zoo dataset.
• Synthetic contains the code for regenerating the experiments on Synthetic dataset.

Please refer to section 5 and 6 of our paper for detailed discussion on experiments.

Execution

The details for executing are mentioned inside each folder. Each folder contains a train.ipynb in which our algorithm and baselines are implemented and the results are illustrated.

Citation

Please cite us if you use our work in your research:

@inproceedings{okati2021differentiable,
      title={Differentiable Learning Under Triage}, 
      author={Nastaran Okati and Abir De and Manuel Gomez-Rodriguez},
      booktitle={Advances in Neural Information Processing Systems},
      year={2021}
}