Pick a loss you can explain on the How to Train Models track. Cross-entropy for classification, squared or absolute error for regression, and ranking losses when order is the product. The loss must be defined on the label you stored.
This lesson assumes you already worked through Train a baseline before a fancy model.
The idea in practice
If labels are probabilities or noisy, say so. Do not use a loss that assumes clean one-hot labels when raters disagree.
A concrete check
goal = {
'track': 'How to Train Models',
'lesson': 'Pick a loss you can explain',
}
checks = [
'input available at decision time',
'score matches the real decision',
'failure case written down',
]
print(goal['lesson'])
for item in checks:
print('-', item)
Run the sketch locally if you have Python. The printout is a reminder of the checks, not a trained model. Replace the strings with the real inputs from your own example before you treat it as a design.
What usually goes wrong
A loss that ignores the rare class you actually care about. When this happens, stop adding parameters or tools. Fix the check, the data, or the permission, then run the same example again.
What to write down
- The input you are allowed to use at decision time.
- The output and the score or pass rule.
- One failure you will test on purpose.
- What you will not claim the system can do.
Practice
Match three tasks to a loss and say why.
Self-check
- Say Pick a loss you can explain in one sentence that mentions an input and an output.
- Name the failure mode in this lesson and the check that would catch it.
Done when: you can explain this lesson without the page open, and you have a written failure case.