Glossary term
Glossary term
Safety and Alignment
Broadly speaking, anything that obscures the signal in a dataset. Noise can be introduced into data in a variety of ways. For example:
Human raters make mistakes in labeling.
Humans and instruments mis-record or omit feature values.
Created for this library
A speech recognition team augments training audio with noise so the model is robust on smartphone and car-cabin microphones.
A computer vision team adds Gaussian noise to training images so the model is robust to low-light conditions.
A research team adds label noise during evaluation to study how the model degrades under realistic annotation errors.
Definition source: Google for Developers Machine Learning Glossary | Creative Commons Attribution 4.0 License