BCQ Perturbation Filter

Problem

Filter a BCQ candidate action by its perturbation distance.

Return action when |action - mu| <= threshold, otherwise fall back to the behavior mean mu.

Hint

BCQ only trusts candidate actions that stay near the behavior policy.

case 1

in: [0.55,0.5,0.1]

expected: 0.55

case 2

in: [0.8,0.5,0.1]

expected: 0.5

case 3

in: [-0.5,0.5,0.2]

expected: 0.5

Starter codePython

def bcq_perturbation_filter(action, mu, threshold):
    # Your code here
    pass

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