Loss as a scoreboard
Loss = surprise; lower is better; it is the game's score.
1The true word was milk. Which model bet more on it, the better predictor?your turn
The cat drinks milk
→ continue← backR replay
Millions of knobs: how do we tune them?
2.4 Gradient descent: rolling downhillBuilds on2.2A first guess: just the last word
Common questions
What is "Loss as a scoreboard" about?
Loss = surprise; lower is better; it is the game's score.
What problem does it solve?
Is this prediction good? By how much?
What will I be able to do after this lesson?
You can explain what training loss measures: the model's surprise at the truth.
What comes next?
Millions of knobs: how do we tune them?