Softmax Function Demo

z₁ (logit 1): 1.0

z₂ (logit 2): 2.0

z₃ (logit 3): 0.5

Softmax(zᵢ) = exp(zᵢ/T) / Σ exp(zⱼ/T)

P(class 1):

33%

P(class 2):

33%

P(class 3):

33%

Temperature: 1.0

T=1 is standard. Lower T → sharper distribution. Higher T → more uniform.