论文标题
部分可观测时空混沌系统的无模型预测
Quantifying the accuracy of steady states obtained from the Universal Lindblad Equation
论文作者
论文摘要
储层计算是预测湍流的有力工具,其简单的架构具有处理大型系统的计算效率。然而,其实现通常需要完整的状态向量测量和系统非线性知识。我们使用非线性投影函数将系统测量扩展到高维空间,然后将其输入到储层中以获得预测。我们展示了这种储层计算网络在时空混沌系统上的应用,该系统模拟了湍流的若干特征。我们表明,使用径向基函数作为非线性投影器,即使只有部分观测并且不知道控制方程,也能稳健地捕捉复杂的系统非线性。最后,我们表明,当测量稀疏、不完整且带有噪声,甚至控制方程变得不准确时,我们的网络仍然可以产生相当准确的预测,从而为实际湍流系统的无模型预测铺平了道路。
We show that steady-state expectation values predicted by the universal Lindblad equation (ULE) are accurate up to bounded corrections that scale linearly with the effective system-bath coupling, $Γ$ (second order in the microscopic coupling). We also identify a near-identity, quasilocal "memory-dressing" transformation, used during the derivation of the ULE, whose inverse can be applied to achieve relative deviations of observables that generically scale to zero with $Γ$, even for nonequilibrium currents whose steady-state values themselves scale to zero with $Γ$. This result provides a solution to recently identified limitations on the accuracy of Lindblad equations, which highlighted a potential for significant relative errors in currents of conserved quantities. The transformation we identify allows for high-fidelity computation of currents in the weak-coupling regime, ensuring thermodynamic consistency and local conservation laws, while retaining the stability and physicality of a Lindblad-form master equation.