MLSqueeze is a tool designed to identify a diversified set of boundary candidates for machine learning classifiers. This helps in better understanding the decision boundaries of classifiers, to support robust and effective testing and validation practice.
- Automated boundary identification for ML and other classifiers.
- Ensures well-diversified boundary candidates.
- Enhances the ability to judge classifier boundaries effectively.
Currently Missing:
- Handling if missing values
- handling of categorical inputs
- Visualization support
TODO
The experiments from the original research paper using MLSqueeze can be found in the dedicated MLSqueeze_sbft_2024 repository.
This project is licensed under the Creative Commons Attribution 4.0 International (CC BY 4.0) License.
For more information, please look into the research paper (https://doi.org/10.1145/3643659.3643927) and cite if you use LMSqueeze in your research:
@inproceedings{dobslaw2024automated,
title={Automated Boundary Identification for Machine Learning Classifiers},
author={Felix Dobslaw and Robert Feldt},
booktitle={2024 ACM/IEEE International Workshop on Search-Based and Fuzz Testing (SBFT '24)},
year={2024},
doi={10.1145/3643659.3643927}
}