Softmax
In plain terms
A function turning scores into a probability distribution.
Going deeper
Softmax converts a vector of numbers into probabilities that sum to one, used for multi-class classification outputs. It highlights the largest values proportionally.
Under the hood
Exponentiates and normalizes; paired with cross-entropy loss for classification.
Examples
Multi-class classification, neural network output, LLM token probabilities.
Think of it like...
Like converting raw vote counts into percentages that add up to 100%.