Compatible Features Check

Problem

Check whether the actor features match the score function gradient, the condition for compatible function approximation. Return True when every element differs by at most tol.

Hint

Compatible features make the policy gradient exact for the linear critic.

Test cases

case 1

in: [[1,2],[1,2],1e-9]

expected: true

case 2

in: [[1,2],[1,2.0000001],0.000001]

expected: true

case 3

in: [[1,2],[1,2.1],0.000001]

expected: false

case 4

in: [[],[],1e-9]

expected: true

Starter codePython

def compatible_features_check(phi, grad_log_pi, tol):
    # Your code here
    pass

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