THE REWARD DISCOUNT LEDGER
Intermediate
Difficulty: Medium Tags: Reinforcement Learning, Returns, Discounting
Story
A reinforcement-learning evaluation service computes discounted returns from a short episode. A reversed loop accidentally discounts the first reward most heavily rather than the last reward.
The Math
For rewards r_0, ..., r_{T-1} and discount gamma in [0, 1], the return at
the start is:
G_0 = sum_{t=0}^{T-1} gamma^t * r_t
Use the reward order as supplied.
Input Format
rewards (length T), gamma.
Output Format
G_0, 6 decimal places.
Constraints
1 <= T <= 10^5,0 <= gamma <= 1,|r_t| <= 10^6.- Time limit: 1.0 second.
Example
Input
rewards = [1, 2, 4], gamma = 0.5
Output
3.000000
Explanation: The return is 1 + 0.5*2 + 0.25*4 = 3.
Keywords
reinforcement-learningreturnsdiscounting
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.