FAIR Digital Object Concept for Composing Machine Learning Training Data

In this poster we introduce how the FAIR¹ Digital Object (FAIR DO) concept can simplify the composition of training data sets for Machine Learning (ML). Training data sets from heterogeneous sources mostly have different label terms. Therefore, composing them for application in ML requires laborious relabeling. To automate this process, the FAIR DO concept can be applied. A FAIR DO is an informative representation of scientific data, e.g. a training data set, that makes the data interpretable and actionable for computer systems. For applicability in the context of ML a FAIR DO requires at least a globally unique Persistent Identifier (PID), mandatory metadata, and a data type. With the self-contained structure of a FAIR DO, the associated label information can be accessed. Here, we show this structure and explain how it facilitates access to label information. Moreover, specialized clients and tools are needed for fully automated acting on FAIR DOs and relabeling. Using FAIR DOs this way could also address other laborious steps in ML training data composition like feature- or file reformatting. The described work is based on the results of the RDA IG FAIR Digital Object Fabric². ... mehrThe outcome contributes to the contents of the RDA IG FAIR for ML³. This work has been supported by the research program ‘Engineering Digital Futures’ of the Helmholtz Association of German Research Centers and the Helmholtz Metadata Collaboration Platform⁴.

(1) FAIR Guiding Principles of scientific data to improve Findability, Accessibility, Interoperability and Reuse of digital assets https://www.go-fair.org/fair-principles/ (2) FAIR Digital Object Fabric IG https://www.rd-alliance.org/group/FAIR-digital-object-fabric-ig.html (3) FAIR for ML https://www.rd-alliance.org/defining-fair-machine-learning-ml (4) Helmholtz Metadata Collaboration (HMC) Platform https://www.hmc-plattform.org/en