凸优化的量子类比
A quantum analogue of convex optimization
AI总结:
本文提出量子基本间隙算法(FGA),在多项式时间内计算具有凸势的薛定谔算子最小特征值,并首次给出求解 n 维凸鼓最低频率的多项式时间量子算法。
摘要:
凸优化是优化理论与实践背后的强大动力。我们引入无约束凸优化的量子类比:计算薛定谔算子 $h = -Δ+ V$ 的最小特征值,其中凸势 $V:\mathbb R^n \rightarrow \mathbb R_{\ge 0}$ 满足当 $\\|x\\|\rightarrow\infty$ 时 $V(x)\rightarrow\infty$。针对该问题,我们提出一种高效的量子算法,称为基本间隙算法(Fundamental Gap Algorithm, FGA),它能够在关于 $n$、$1/ε$ 以及依赖于 $V$ 的参数的多项式时间内,以误差 $ε$ 计算 $h$ 的最小特征值。基态的绝热演化被用作关键子程序,我们利用新颖的技术对其进行分析,从而能够聚焦于低能空间。我们将 FGA 应用于给出首个已知的多项式时间算法,用于寻找 $n$ 维凸鼓的最低频率,数学上即在由 $m$ 个线性约束定义的 $n$ 维区域上 Dirichlet 拉普拉斯算子的最小特征值,其运行时间关于 $n$、$m$、$1/ε$ 以及包围该区域的球的半径 $R$ 是多项式的。
英文摘要:
Convex optimization is the powerhouse behind the theory and practice of optimization. We introduce a quantum analogue of unconstrained convex optimization: computing the minimum eigenvalue of a Schrödinger operator $h = -Δ+ V $ with convex potential $V:\mathbb R^n \rightarrow \mathbb R_{\ge 0}$ such that $V(x)\rightarrow\infty $ as $\|x\|\rightarrow\infty$. For this problem, we present an efficient quantum algorithm, called the Fundamental Gap Algorithm (FGA), that computes the minimum eigenvalue of $h$ up to error $ε$ in polynomial time in $n$, $1/ε$, and parameters that depend on $V$. Adiabatic evolution of the ground state is used as a key subroutine, which we analyze with novel techniques that allow us to focus on the low-energy space. We apply the FGA to give the first known polynomial-time algorithm for finding the lowest frequency of an $n$-dimensional convex drum, or mathematically, the minimum eigenvalue of the Dirichlet Laplacian on an $n$-dimensional region that is defined by $m$ linear constraints in polynomial time in $n$, $m$, $1/ε$ and the radius $R$ of a ball encompassing the region.
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。