Open Science practices
CORENET aims to maximise the impact of its results while following the EU’s Open Science policy. To this end, sharing of publications and data will be based on the following practices:
Open data sharing
Supporting the 8 ambitions of the EU’s open science policy and aligned with the EC data management rules, all efforts will be made to ensure open sharing of research data, following provisions set in the DMP.
Open-source and open methodology
Most of Watson Machine Learning Tools used at IBM are based on open-source and combined with IBM software for code-based and visual data science (e.g., PyTorch, TensorFlow and scikit-learn). IBM puts tremendous efforts in explainable AI and developed processes and methods that allow human users to comprehend and trust the results and output created by AI algorithms, including its expected impact and potential biases. This is for instance such an add-on to open-source machine learning algorithms.
The tools developed at SDU will be integrated into their own open-source generative chemistry framework, all sources will be released on GitHub, and a live server will be provided to the community for easy testing. Key for this are already available in-house open-source libraries for state-of-the-art graph canonicalization, specialised open-source libraries like xtb for semiempirical quantum mechanical approaches or DeepChem (an open-source toolchain for the use of deep-learning in materials science, quantum chemistry, and biology), as well as highly efficient and flexible core libraries as provided by boost or mkl.
Thus, all tools developed in CORENET will build on existing open-source software and will be made freely available via GitHub.
Open access to publications
In line with the Horizon Europe guidelines, all efforts will be made to ensure the sharing of results via open access peer-reviewed publications, preferentially following the gold open access route. All publications will be deposited in a trusted repository, such as Zenodo, and will be shared via the project website and through the partners’ networks.