2 ms·
You can hardly train RL in the real world (as opposed to simulation) because of its high sample complexity.
by visarga 8y ago
You can hardly train RL in the real world (as opposed to simulation) because of its high sample complexity.
- red75prime 8y agoExacerbated by sparse rewards. Well, it is being tackled as everything else. [0] for example. [0] https://arxiv.org/pdf/1802.10567.pdf https://arxiv.org/pdf/1802.10567.pdf