Artificial intelligence · Advanced university
A two-score softmax probability
Softmax normalizes scores; this alone does not prove probabilities are calibrated to real-world frequencies.
Method and conditions
Softmax normalizes scores; this alone does not prove probabilities are calibrated to real-world frequencies.
p₁ = exp(s₁)/(exp(s₁)+exp(s₂))
Worked examples
Softmax scores are (-2,1). Calculate the probability assigned to the first score. Round the final result to 4 decimal places if needed.
Use the stable equivalent 1/[1+exp(3)].
0.0474
Softmax scores are (0,-4). Calculate the probability assigned to the first score. Round the final result to 4 decimal places if needed.
Use the stable equivalent 1/[1+exp(-4)].
0.982
Try it
Softmax scores are (3,3). Calculate the probability assigned to the first score. Round the final result to 4 decimal places if needed.
Softmax scores are (2,4). Calculate the probability assigned to the first score. Round the final result to 4 decimal places if needed.
Check your work
Answers and marking notes
- 0.5
Use the stable equivalent 1/[1+exp(0)].
- 0.1192
Use the stable equivalent 1/[1+exp(2)].
Further reading and review references
These are reference links, not copied exercises, institutional endorsement or proof of independent review.
AI-assisted preview; independent educational review required. Level labels describe prerequisites, not accreditation or official curriculum alignment.