Errors and bugs can easily slip in the code during development. In order to enhance code quality, techniques from standard programming – such as peer review – can also be applied to machine learning code.
Peer review is a well known technique, in which members of the team review the code between themselves. This technique is known to help to:
- avoid errors,
- ease debugging,
- ease incident response,
- enhance knowledge transfer within the team.
Whenever possible, use a collaborative software development platform (e.g. Github), which already has mature features for peer review.