Calibration on the How to Train Models track. A probability of 0.8 should be right about 80 percent of the time. Many models rank well and are badly calibrated. If you threshold probabilities, check a reliability curve.
This lesson assumes you already worked through Evaluate after you stop.
The idea in practice
Bin predictions and compare confidence to accuracy. Recalibrate on validation if you must, never on test.
A concrete check
goal = {
'track': 'How to Train Models',
'lesson': 'Calibration',
}
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
Treating a softmax score as a true probability with no check. 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
Explain what a reliability curve compares.
Self-check
- Say Calibration 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.