Loss: the score training reduces on the AI Learning track. Loss is a number that is high when the prediction is wrong and low when it is right. Training tries to lower the average loss. The loss is not the business metric unless you chose it to be.
This lesson assumes you already worked through Labels and who makes them.
The idea in practice
Classification often uses cross-entropy. Regression often uses squared error. If you care about ranking, a classification loss may be the wrong score. Say which decision the loss ignores.
A concrete check
goal = {
'track': 'AI Learning',
'lesson': 'Loss: the score training reduces',
}
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
Celebrating a lower loss while the business metric, such as false alarms, gets worse. 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
Name the loss you would use to predict a price, and one business number you would still report separately.
Self-check
- Say Loss: the score training reduces 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.