Tyler Carter
-
What is your workflow for auditing potential data leakage inside your experi...
Method Help
Data leakage is a pervasive but often silent problem. Information from test set accidentally leaks into training procedure and produces over‑optimistic evaluation results. What step‑by‑step audit routine helps you hunt for leakage risks?
Feng Song, Feng Shi and 32 others5 CommentsView more comments-
Trace every data transformation step and verify no test‑set information is used during training phase.
2 -
Inspect preprocessing scripts: scaling, normalisation must be fit strictly on training split only.
2 - View more comments
-
