How It Works
Training / Training Data
The process (and the data used) to teach a model patterns by showing it many examples before it’s used to make predictions.
Origin · no single documented coiner
Standard machine-learning vocabulary with no single documented coiner. The push to document and audit what is actually in training data is more traceable: Timnit Gebru and Kate Crawford proposed datasheets for datasets in 2018, and Abeba Birhane has since published audits finding racist, misogynistic and non-consensual material inside widely used open datasets. All three are profiled on our Who’s Who page.
No single primary source link for this term.