Thartvigsen / StopAndHop

Code for Stop&Hop, a method for learning to classify irregularly-sampled time series early
17 stars 1 forks source link

Stop&Hop: Early Classification of Irregular Time Series Github_Picture

This repository includes all necessary components to replicate our experiments and use our proposed method.

Working examples

We provide two examples for using Stop&Hop:

Pretrained Embeddings

First, we provide dataloaders from an RNN pretrained on the physionet dataset, which can be found in pretrained_example.py. To run this example, you can create a virtual environment from our requirements.txt file:

python3 -m venv stophop_venv

source stophop_venv/bin/activate

Joint Learning for RNN and HaltingPolicy

Second, we provide access to our ExtraSensory datasets, which can be run using example.py.

Download our data

You can download our data here, and put them into a directory called data.

Citation

Please use the following to cite this work:

@inproceedings{hartvigsen2022stophop,
  title={Stop\&Hop: Early Classification of Irregular Time Series},
  author={Hartvigsen, Thomas and Gerych, Walter and Thadajarassiri, Jidapa and Kong, Xiangnan and Rundensteiner, Elke},
  booktitle={Proceedings of the 31st ACM International Conference on Information & Knowledge Management},
  year={2022}
}