2.3.2 Sufficiency of the learning process
Assurance objective: The learning approach is appropriate to satisfy the defined safety requirements.
Contextual description: The approach taken to learning can affect assurance in a number of ways. The types of models, templates and parameters selected by the system developers during the machine learning process can all impact the satisfaction of the requirements. Decisions are also required on whether an off-line or on-line learning approach is most appropriate. There will also be uncertainty in the outputs of the learned algorithms (such as a stated confidence in a classification) that must be accounted for.