scikit-weak 0.1.4a1

A package featuring utilities and algorithms for weakly supervised ML.

# scikit-weak (scikit-weakly-supervised)

A package featuring utilities and algorithms for weakly supervised ML. Should be (more-or-less) compatible with scikit-learn! It collects original algorithms and methods developed at the MUDI lab (DISCo dept., University of Milano-Bicocca, Milan, Italy), as well as some algorithms available in the literature.

## How to install You can install the library using the command:

` pip install scikit-weak `

### Dependencies: numpy, scipy, scikit-learn, pandas

## Documentation The documentation is generated using Sphinx (https://www.sphinx-doc.org/). If you download the source code from this repository you can generate the documentation in html format by typing: ` sphinx-build -b html docs/source docs/build/html ` in the main folder of the project.

## References:

[1] Campagner, A., Ciucci, D., Hullermeier, E. (2021). Rough set-based feature selection for weakly labeled data. International Journal of Approximate Reasoning, 136, 150-167. https://doi.org/10.1016/j.ijar.2021.06.005.

[2] Campagner, A., Ciucci, D., Svensson, C. M., Figge, M. T., & Cabitza, F. (2021). Ground truthing from multi-rater labeling with three-way decision and possibility theory. Information Sciences, 545, 771-790. https://doi.org/10.1016/j.ins.2020.09.049

[3] Campagner, A., Ciucci, D., & Hüllermeier, E. (2020). Feature Reduction in Superset Learning Using Rough Sets and Evidence Theory. In International Conference on Information Processing and Management of Uncertainty in Knowledge-Based Systems (pp. 471-484). Springer, Cham. https://doi.org/10.1007/978-3-030-50146-4_35