방법 비교
선택한 방법을 나란히 검토하세요. 서로 다른 행은 강조 표시됩니다.
| 약한 지도 강화학습× | 자기 지도 강화 학습× | |
|---|---|---|
| 분야 | 딥러닝 | 딥러닝 |
| 계열 | Machine learning | Machine learning |
| 기원 연도≠ | 2010s–present | 2020 |
| 창시자≠ | Multiple contributors; reward-learning framing: Christiano et al. (2017) | Laskin, M.; Srinivas, A.; Abbeel, P. (and contemporaries) |
| 유형≠ | Reinforcement learning with imperfect or partial reward supervision | Self-supervised auxiliary-task learning for RL |
| 원전≠ | Sutton, R. S. & Barto, A. G. (2018). Reinforcement Learning: An Introduction (2nd ed.). MIT Press. ISBN: 978-0-262-03924-6 | Laskin, M., Srinivas, A., & Abbeel, P. (2020). CURL: Contrastive Unsupervised Representations for Reinforcement Learning. Proceedings of the 37th International Conference on Machine Learning (ICML), PMLR 119, 5639–5650. link ↗ |
| 별칭 | WSRL, weak-reward RL, imperfect-reward reinforcement learning, reward-impoverished RL | SSL-RL, self-supervised RL, representation-based reinforcement learning, auxiliary-task RL |
| 관련≠ | 3 | 4 |
| 요약≠ | Weakly supervised reinforcement learning (WSRL) trains agents in environments where the reward signal is imperfect, sparse, delayed, or only partially informative — unlike dense fully-supervised RL. The agent must learn effective policies despite incomplete feedback, using auxiliary signals, reward modeling, or preference learning to compensate for the weak supervision. | Self-supervised Reinforcement Learning (SSL-RL) augments standard RL training with self-supervised auxiliary objectives — such as contrastive, predictive, or data-augmentation-based tasks — applied to the agent's own experience. These objectives improve the quality of learned representations without requiring extra human labels, enabling faster convergence and better sample efficiency, especially in high-dimensional observation spaces like raw pixels. |
| ScholarGate데이터셋 ↗ |
|
|