arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2103.09847cs.LGcs.AImath.OCstat.ML
收藏

线性函数逼近下的无限时域离线强化学习:收敛性诅咒与算法

Infinite-Horizon Offline Reinforcement Learning with Linear Function Approximation: Curse of Dimensionality and Algorithm

University of California, Berkeley(加州大学伯克利分校);Université de Lorraine(洛林大学);Inria(法国国家信息与自动化研究所);IECL(IECL研究所)

Lin Chen, Bruno Scherrer, Peter L. Bartlett

摘要:

在本文中,我们研究的样本复杂性的政策评估在无限时域离线强化学习(也称为离政策评估问题)与线性函数逼近。我们确定了一个硬政权$dγ^{2}>1$,其中$d$是特征向量的维度和$γ$是折扣率。在这种情况下,对于任何$q\in[γ^{2},1]$,我们可以构造一个硬实例,使得其特征协方差矩阵的最小特征值是$q/d$,并且它需要$Ω\left(\frac{d}{γ^{2}\left(q-γ^{2}\right)\varepsilon^{2}}\exp\left(Θ\left(dγ^{2}\right)\right)\right)$样本来近似值函数直到加性误差$\varepsilon$。注意,样本复杂度的下限在$d$中是指数的。如果是$q=γ^{2}$,即使无限的数据也不够。低分布移位的假设下,我们表明,有一个算法,需要在最多的$O\left(\max\left\{ \frac{\left\Vert θ^π\right\Vert _{2}^{4}}{\varepsilon^{4}}\log\frac{d}δ,\frac{1}{\varepsilon^{2}}\left(d+\log\frac{1}δ\right)\right\} \right)$样本($θ^π$是参数的政策,在线性函数逼近),并保证近似值函数的添加剂错误的$\varepsilon$的概率至少$1-δ$。

英文摘要:

In this paper, we investigate the sample complexity of policy evaluation in infinite-horizon offline reinforcement learning (also known as the off-policy evaluation problem) with linear function approximation. We identify a hard regime $dγ^{2}>1$, where $d$ is the dimension of the feature vector and $γ$ is the discount rate. In this regime, for any $q\in[γ^{2},1]$, we can construct a hard instance such that the smallest eigenvalue of its feature covariance matrix is $q/d$ and it requires $Ω\left(\frac{d}{γ^{2}\left(q-γ^{2}\right)\varepsilon^{2}}\exp\left(Θ\left(dγ^{2}\right)\right)\right)$ samples to approximate the value function up to an additive error $\varepsilon$. Note that the lower bound of the sample complexity is exponential in $d$. If $q=γ^{2}$, even infinite data cannot suffice. Under the low distribution shift assumption, we show that there is an algorithm that needs at most $O\left(\max\left\{ \frac{\left\Vert θ^π\right\Vert _{2}^{4}}{\varepsilon^{4}}\log\frac{d}δ,\frac{1}{\varepsilon^{2}}\left(d+\log\frac{1}δ\right)\right\} \right)$ samples ($θ^π$ is the parameter of the policy in linear function approximation) and guarantees approximation to the value function up to an additive error of $\varepsilon$ with probability at least $1-δ$.

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。

↑