Constrained Policy Update

Problem

Apply a Lagrangian penalty to advantages: A_pen[a] = advantages[a] - lambda * cost_advantages[a]. Return the penalized advantage list.

Hint

The multiplier trades reward against constraint violation.

case 1

in: [[1,2],[0.5,1],1]

expected: [0.5,1]

case 2

in: [[0],[2],0.5]

expected: [-1]

case 3

in: [[],[],1]

expected: []

Starter codePython

def constrained_advantage_update(advantages, cost_advantages, lambda_):
    # Your code here
    pass

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