AI & MLCoreAIML-ATT-003
Scaling and Softmax
Scores into weights: why divide by root-d, and the sum-to-one guarantee.
- Difficulty
- 2 / 5
- Importance
- 5 / 5
- Study time
- 1–2 h
- Kind of work
- Mixed
Why it matters
The learner can scale and softmax tiny scores and explain what unscaled growth would do to gradients.