DQN playing LunarLander-v2 from https://github.com/sgoodfriend/rl-algo-impls/tree/1d4094fbcc9082de7f53f4348dd4c7c354152907
9839b09
| import numpy as np | |
| def discounted_cumsum(x: np.ndarray, gamma: float) -> np.ndarray: | |
| dc = x.copy() | |
| for i in reversed(range(len(x) - 1)): | |
| dc[i] += gamma * dc[i + 1] | |
| return dc | |