One number to optimize, a human yardstick to beat, and an error analysis that says what to do next.
9h · 12 lessons · 1 challenge
Before this: Optimization Algorithms & Tuning
This module is about the decisions around the model rather than the model. It opens where the source course does: orthogonalization, and choosing a single number to optimize so that two candidates can actually be compared.
Then human-level performance as the yardstick that turns error into avoidable bias, error analysis done by counting, what to do when training and deployment data differ, and when transfer or multi-task learning is the cheaper answer.
Six proposed fixes and one error analysis. Say what each is worth before you are told, then see the ceiling it actually removes.
Worth 450 XP.
