Artificial Intelligence – Uncertainty quantification in machine learning; Text in English

Scope

This DIN SPEC has been developed within a DIN SPEC consortium. The document was developed and approved by organizations named in the Foreword. This document specifies general guidance and requirements for the development and use of methods for quantifying uncertainty in machine learning (ML). This document defines fundamental terminology for uncertainty quantification for ML and specifies the purpose, use and necessity for these quantifications. The document provides an overview of available uncertainty quantification approaches in ML and their characteristics, and describes selected applications. This document specifies general technical requirements and recommendations for uncertainty quantification for ML. This document is applicable to research, development, and the usage of ML techniques including but not limited to medicine, earth observation, natural language processing (NLP), finance, computer vision, autonomous mobility and machinery. This document does not specify requirements for uncertainty criteria necessary for safety functions and health protection, such as for the detection of persons in a hazardous area. This also applies to so-called assistance systems. © 2024 DIN Media GmbH

External Links