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 2602.14342math.STcs.DScs.LGmath.PRstat.TH
收藏

高精度对数凹采样与随机查询

High-accuracy log-concave sampling with stochastic queries

MIT(麻省理工学院);Yale University(耶鲁大学)

Fan Chen, Sinho Chewi, Constantinos Daskalakis, Alexander Rakhlin

AI总结:

本文提出利用具有亚指数尾部的随机梯度实现高精度对数凹采样,展示其在迭代和查询复杂度上与凸优化问题的分离,并提供信息论论证证明轻尾随机梯度对高精度的必要性。

摘要:

我们证明,利用具有亚指数尾部的随机梯度,可以实现对数凹采样的高精度保证,即迭代和查询复杂度随$\mathrm{poly}\log(1/δ)$增长,其中$δ$是目标精度。值得注意的是,这与凸优化问题形成对比,后者即使在梯度 oracle 中存在加性高斯噪声,也会导致$\mathrm{poly}(1/δ)$次查询。我们还提供了一个信息论论证,证明轻尾随机梯度对于高精度的必要性:例如在方差有界的情况下,我们展示最小最大最优查询复杂度随$Θ(1/δ)$增长。我们的框架也提供了在随机零阶(值)查询下的类似高精度保证,并为从有限和势能采样提供了改进的复杂度结果。

英文摘要:

We show that high-accuracy guarantees for log-concave sampling -- that is, iteration and query complexities which scale as $\mathrm{poly}\log(1/δ)$, where $δ$ is the desired target accuracy -- are achievable using stochastic gradients with subexponential tails. Notably, this exhibits a separation with the problem of convex optimization, where stochasticity (even additive Gaussian noise) in the gradient oracle incurs $\mathrm{poly}(1/δ)$ queries. We also give an information-theoretic argument that light-tailed stochastic gradients are necessary for high accuracy: for example, in the bounded variance case, we show that the minimax-optimal query complexity scales as $Θ(1/δ)$. Our framework also provides similar high accuracy guarantees under stochastic zeroth order (value) queries, and an improved complexity result for sampling from finite-sum potentials.

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。

↑