Article type
Abstract
Background and objective
Scientific machine learning (SML) is a cross-cutting research area combining physics-based models, machine learning, and artificial intelligence in the domains of engineering and physical sciences. The rapid development of SML methods and the use of ever more complex tools create a tangible problem of model interpretation, transparency, reproducibility, and validation. Many existing publications fail to report key information, hindering usability and accessibility. Although reporting guidelines are the norm in health sciences, this is not the case for physical sciences. We propose the development of a reporting guideline to aid authors to thoroughly describe all elements of an SML-based study.
Methods
Guideline development requires participation of a broad network of scientists, researchers, and other key stakeholders and is a formal process based on a protocol. The guideline comprises a set of items concerning reporting and quality recommendations presented in the form of a checklist. This checklist will address 2 main areas: the data used (evidence base) and the methods used (models). This overarching methodology is based on health research.
The people participating in the process can be categorized into 4 groups, which are not mutually exclusive. The executive committee will be tasked with overall organization of the project. A Delphi panel will be used to identify a list of potential items to be included in the guideline, based on a ranking-type questionnaire. An advisory group will be formed by participants of a face-to-face consensus meeting, where the guideline will be formally discussed and consensus reached. Finally, a writing group will undertake formal writing of the guideline as a main guideline and an accompanying explanatory document. These are not usual groupings in SML, so there will be novelty in even establishing such a methodology.
Conclusions
The guideline will be a checklist of the minimal items that authors should report, while the explanatory document will go into further detail providing applied examples of good-quality reporting. Authors, editors, peer reviewers, and funding bodies can benefit as the guideline will help not only the writing of papers but also peer reviewing and funded research assessment.
Scientific machine learning (SML) is a cross-cutting research area combining physics-based models, machine learning, and artificial intelligence in the domains of engineering and physical sciences. The rapid development of SML methods and the use of ever more complex tools create a tangible problem of model interpretation, transparency, reproducibility, and validation. Many existing publications fail to report key information, hindering usability and accessibility. Although reporting guidelines are the norm in health sciences, this is not the case for physical sciences. We propose the development of a reporting guideline to aid authors to thoroughly describe all elements of an SML-based study.
Methods
Guideline development requires participation of a broad network of scientists, researchers, and other key stakeholders and is a formal process based on a protocol. The guideline comprises a set of items concerning reporting and quality recommendations presented in the form of a checklist. This checklist will address 2 main areas: the data used (evidence base) and the methods used (models). This overarching methodology is based on health research.
The people participating in the process can be categorized into 4 groups, which are not mutually exclusive. The executive committee will be tasked with overall organization of the project. A Delphi panel will be used to identify a list of potential items to be included in the guideline, based on a ranking-type questionnaire. An advisory group will be formed by participants of a face-to-face consensus meeting, where the guideline will be formally discussed and consensus reached. Finally, a writing group will undertake formal writing of the guideline as a main guideline and an accompanying explanatory document. These are not usual groupings in SML, so there will be novelty in even establishing such a methodology.
Conclusions
The guideline will be a checklist of the minimal items that authors should report, while the explanatory document will go into further detail providing applied examples of good-quality reporting. Authors, editors, peer reviewers, and funding bodies can benefit as the guideline will help not only the writing of papers but also peer reviewing and funded research assessment.