A semi-automated pipeline to enable trustworthy AI. We intend to address the problem of realizing a trustworthy system using technical methods.
The following figure shows the proposed workflow to achieve a trustworthy AI. The "Diabetes Dataset" and "Heart Disease Dataset" folders contain an example with different datasets of the methodology. For each dataset, there is a notebook per requirement so that anyone can consult the practical application of the methodology.
Methods for achieving the trustworthy AI requirements across all lifecycle phases.
- Download the whole repository.
- Choose one dataset.
- Look at the methodology proposed for each requirement in its single notebook
.ipynb. - Apply the methods to your specific dataset.
It is only necessary to modify and adapt the Data Collection and Metadata script in order to incorporate specific information.
If you use this code please cite:
Carlos de Manuel, David Fernández-Narro, Vicent Blanes-Selva, Juan M García-Gómez, Carlos Sáez. A Development Framework for Trustworthy Artificial Intelligence in Health with Example Code Pipelines.
- Version: 1.0.0
- Authors: Carlos de Manuel Vicente (UPV), David Fernández-Narro (UPV), Vicent Blanes-Selva (UPV), Juan M García-Gómez (UPV), Carlos Sáez (UPV).
Copyright: 2024 - Biomedical Data Science Lab, Universitat Politècnica de València, Spain (UPV)
If you are interested in collaborating in this work please contact us.