'Is it possible to make the reinforcement learning agent output 4D actions?
Is it possible to make the agent output 4D actions? I mean, there are three actions possible (0, 1, 2), but they should be combined in four dimensions.
ex) Context-bandit situation
input_t1: [0, 1, 2, 0]
input_t2: [2, 0, 1, 0]
output: [1, 2, 0, 0]
Solution 1:[1]
Yes, just use MultiDiscrete actions (from gym.spaces)
self.action_space = MultiDiscrete([number_of_possible_actions for _ in range(self.number_of_action_dimensions)])
Sources
This article follows the attribution requirements of Stack Overflow and is licensed under CC BY-SA 3.0.
Source: Stack Overflow
| Solution | Source |
|---|---|
| Solution 1 | Raoul Raftopoulos |