Lab 4: threshold optimization on a model that cannot be retrained — Hands-on Labs — Atlas