Double DQN: Measuring Overestimation
Beginner
The problem, from first principles
The Double DQN paper measures how much the standard target overestimates. Comparing the two targets for the same transition gives a direct check on the bias the method removes.
From theory to code
Implement overestimation_gap(max_target_q, double_target), returning their difference.
Constraints
- Returns a Python float.
Hints
Hint 1
Subtract the Double DQN target from the max target.
Keywords
research-papersreinforcement-learningdeep-q-learning
Python
Run your code to see results
No results yet
Run your code to check it against the examples, or submit to run every test.