揭示了动作表征学习中为何离散表示更有效
On the Identifiability of Latent Action Policies
- 提出可识别性分析框架，定义理想动作表征标准
- 证明熵正则化目标能确保满足理想标准的表征
- 解释实践中离散动作表征表现优异的原因
我们研究了最近提出的从视频数据中发现动作表征的潜在动作策略学习（LAPO）框架的可识别性。本文形式化描述了此类表征的理想特性、其统计优势以及可能导致不可识别性的来源。最终证明，在适当条件下，熵正则化的LAPO目标能够识别出满足我们理想标准的动作表征。该分析为离散动作表征在实践中表现良好提供了理论解释。
原文摘要 · Abstract (English)
We study the identifiability of latent action policy learning (LAPO), a framework introduced recently to discover representations of actions from video data. We formally describe desiderata for such representations, their statistical benefits and potential sources of unidentifiability. Finally, we prove that an entropy-regularized LAPO objective identifies action representations satisfying our desiderata, under suitable conditions. Our analysis provides an explanation for why discrete action representations perform well in practice.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。