arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2012.01491cs.LG
收藏

策略梯度求二阶平稳点的样本复杂性

Sample Complexity of Policy Gradient Finding Second-Order Stationary Points

Zhejiang University(浙江大学);Nanyang Technological University(南洋理工大学)

Long Yang, Qian Zheng, Gang Pan

摘要:

基于策略的强化学习(RL)的目标是搜索其目标的最大值点。然而,由于其目标的内在非线性,收敛到一阶稳定点(FOSP)不能保证策略梯度方法找到一个最大点。一个FOSP可以是一个最小值,甚至是一个鞍点,这对于RL来说是不可取的。幸运的是,如果所有的鞍点都是\emph{strict},所有的二阶平稳点(SOSP)完全等价于局部极大值。我们用SOSP代替FOSP作为表征策略梯度样本复杂性的收敛准则。我们的结果表明,策略梯度在总代价为$\mathcal{O}\left(\dfrac{ε^{-\frac{9}{2}}}{(1-γ)\sqrtχ}\log\dfrac{1}δ\right)$之后以至少$1-\widetilde{\mathcal{O}}(δ)$的概率收敛到$(ε,\sqrt{εχ})$ -SOSP,其中$γ\in(0,1)$ .我们的结果显著改善了需要$\mathcal{O}\left(\dfrac{ε^{-9}χ^{\frac{3}{2}}}δ\log\dfrac{1}{εχ}\right)$的最新结果。我们的分析是基于一个关键的思想,即分解的参数空间$\mathbb{R}^p$到三个非平稳区域:非平稳点,鞍点,局部最优区域,然后在每个区域的RL的目标进行局部改进。这种技术可以潜在地推广到广泛的策略梯度方法。

英文摘要:

The goal of policy-based reinforcement learning (RL) is to search the maximal point of its objective. However, due to the inherent non-concavity of its objective, convergence to a first-order stationary point (FOSP) can not guarantee the policy gradient methods finding a maximal point. A FOSP can be a minimal or even a saddle point, which is undesirable for RL. Fortunately, if all the saddle points are \emph{strict}, all the second-order stationary points (SOSP) are exactly equivalent to local maxima. Instead of FOSP, we consider SOSP as the convergence criteria to character the sample complexity of policy gradient. Our result shows that policy gradient converges to an $(ε,\sqrt{εχ})$-SOSP with probability at least $1-\widetilde{\mathcal{O}}(δ)$ after the total cost of $\mathcal{O}\left(\dfrac{ε^{-\frac{9}{2}}}{(1-γ)\sqrtχ}\log\dfrac{1}δ\right)$, where $γ\in(0,1)$. Our result improves the state-of-the-art result significantly where it requires $\mathcal{O}\left(\dfrac{ε^{-9}χ^{\frac{3}{2}}}δ\log\dfrac{1}{εχ}\right)$. Our analysis is based on the key idea that decomposes the parameter space $\mathbb{R}^p$ into three non-intersected regions: non-stationary point, saddle point, and local optimal region, then making a local improvement of the objective of RL in each region. This technique can be potentially generalized to extensive policy gradient methods.

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。

↑