bandit.algo.softmax — bandit/bandit-core 0.2.0

Softmax algorithm

p clj

(p temperature z value)

probabilities clj

(probabilities temperature arms)

probability clj

(probability temperature {:keys [value] :as arm} arms)

select-arm clj

(select-arm temperature arms)

(select-arm temperature rand-val arms)

selects the arm to pull. a higher temperature causes the algorithm to favour experimentation. 0 < temperature < 1.

z clj

(z temperature values)