Potential Shaping Reward

Problem

Compute the potential-based shaping reward F(s, s') = gamma * phi(s') - phi(s). This form provably leaves the optimal policy unchanged.

Hint

Any shaping shaped like a discounted potential difference preserves optimal policies.

case 1

in: [1,2,0.9]

expected: 0.8

case 2

in: [0,0,0.9]

expected: 0

case 3

in: [5,1,0.5]

expected: -4.5

case 4

in: [-2,3,1]

expected: 5

Starter codePython

def potential_shaping_reward(phi_s, phi_s_next, gamma):
    # Your code here
    pass

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