A package for a score-based change detection method, auto-test, that can automatically report hidden changes in machine learning systems as they learn from a continuous, possibly evolving, stream of data. This is code accompanying the paper "Score-Based Change Detection for Gradient-Based Learning Machines" in ICASSP 2021.
This package is based on PyTorch. Other dependencies can be found in the file environment.yml.
If using conda, run the following command to install all required packages and activate the environment:
$ conda env create --file environment.yml
$ source activate autodetectClone the repository here:
$ git clone https://github.com/langliu95/autodetect.git
$ cd autodetect/The documentation for this package can be found here.
- Lang Liu
- Joseph Salmon
- Zaid Harchaoui
This project is licensed under the GPLv3 License - see the LICENSE file for details.
If you use this code, please cite:
@inproceedings{lsh2021,
title = {Score-Based Change Detection for Gradient-Based Learning Machines},
author = {Liu, Lang and
Salmon, Joseph and
Harchaoui, Zaid},
booktitle = {2021 {IEEE} International Conference on Acoustics, Speech and Signal Processing, {ICASSP} 2021, Toronto, Canada, June 6-11, 2021},
publisher = {{IEEE}},
year = {2021}
}
This work was supported by NSF CCF-1740551, NSF DMS-1810975, the program “Learning in Machines and Brains” of CIFAR, and faculty research awards.