Behavior Policy Estimate

Problem

Estimate the behavior policy from action counts.

Return counts[a] / total for each action. If all counts are zero return zeros, and an empty list returns [].

Hint

Offline methods often need the behavior propensity of each logged action.

case 1

in: [[3,1]]

expected: [0.75,0.25]

case 2

in: [[0,0]]

expected: [0,0]

case 3

in: [[2,2,2,2]]

expected: [0.25,0.25,0.25,0.25]

case 4

in: [[]]

expected: []

Starter codePython

def behavior_policy_estimate(action_counts):
    # Your code here
    pass

4 test cases · run your code against them in the editor