Invalid Action Penalty

Problem

Penalize an action that is not in the valid action list.

Return reward if action is valid, otherwise reward - penalty.

Hint

This is a soft alternative to hard action masking.

case 1

in: [1,2,[0,1,2],5]

expected: 1

case 2

in: [1,3,[0,1,2],5]

expected: -4

case 3

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

expected: -2

case 4

in: [-1,1,[1],0.5]

expected: -1

Starter codePython

def invalid_action_penalty(reward, action, valid_actions, penalty):
    # Your code here
    pass

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