scikit-weak is a Python library for weakly supervised learning, compatible with (and based on) scikit-learn. I'm the primary developer and mantainer of the library, which also feature contributions by Julian Lienen from Paderborn University. The library is open-source and available on GitHub and PyPi. The library currently focuses on learning from imprecise data, self-supervised learning and label regularization problems, by providing a framework for management and representation of weakly supervised data, state-of-the-art classification algorithms, as well as pre-processing, data management, and evaluation utilities.

DSS Quality Assessment

DSS Quality Assessment is a Web Tool for the assessment of data-driven decision support systems, especially those based on AI and ML methodologies. The Web Tool encompassess different functionalities to provide a holistic approach to quality assessment, in particular it allows to: evaluate the similarity of datasets, evaluate the calibration and robustness of decision support systems, as well as evaluate the utility and impact of a decision support system on human decision-making. The web tool was developed at the MUDI Lab of the University of Milano-Bicocca and draws from multiple years of research conducted at the same lab. The web tool is available here.

COVID-19 Blood-ML

The aim of this project was to develop diagnostic and prognostic models for COVID-19 disease, based on blood test data. The project has led to the collection of multiple datasets from several hospitals and clinics around 4 countries (Italy, Spain, Brazil, Ethiopia), which are publicly available (for benchmarking, analysis and model development purposes) on Zenodo: The developed ML diagnostic models have been deployed through freely usable web-tools COVID-19-BLOOD-ML and COVID19-BLOODTESTS, or as standalone models on GitHub.