arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2105.01136stat.MLcs.LG
收藏

通过张量分解学习良好的状态和动作表示

Learning Good State and Action Representations via Tensor Decomposition

Princeton University(普林斯顿大学);Massachusetts Institute of Technology(麻省理工学院);Duke University(杜克大学)

Chengzhuo Ni, Yaqi Duan, Munther Dahleh, Anru Zhang, Mengdi Wang

摘要:

连续状态作用马尔可夫决策过程(MDP)的转移核具有自然张量结构。本文提出了一种张量启发的无监督学习方法,从经验轨迹中识别有意义的低维状态和动作表示。该方法通过核化、重要性抽样和低Tucker秩近似来利用MDP的张量结构。该方法可以进一步用于状态和动作的聚类,并找到最佳的离散MDP抽象。我们提供了尖锐的统计误差范围张量浓度和保存后嵌入的扩散距离。我们进一步证明,学习的状态/动作抽象提供准确的近似潜在的块结构,如果它们存在,使功能近似在下游任务,如政策评估。

英文摘要:

The transition kernel of a continuous-state-action Markov decision process (MDP) admits a natural tensor structure. This paper proposes a tensor-inspired unsupervised learning method to identify meaningful low-dimensional state and action representations from empirical trajectories. The method exploits the MDP's tensor structure by kernelization, importance sampling and low-Tucker-rank approximation. This method can be further used to cluster states and actions respectively and find the best discrete MDP abstraction. We provide sharp statistical error bounds for tensor concentration and the preservation of diffusion distance after embedding. We further prove that the learned state/action abstractions provide accurate approximations to latent block structures if they exist, enabling function approximation in downstream tasks such as policy evaluation.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑