Value Function Loss

Problem

Compute the mean squared error between predicted state values and their targets. An empty batch returns 0.0.

Hint

The critic regresses the value estimate onto the bootstrapped return target.

case 1

in: [[1,2],[1,4]]

expected: 2

case 2

in: [[],[]]

expected: 0

case 3

in: [[3],[3]]

expected: 0

case 4

in: [[0,1,2],[2,1,0]]

expected: 2.6666666666666665

Starter codePython

def value_loss(values, targets):
    # Your code here
    pass

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