论文标题
部分可观测时空混沌系统的无模型预测
Descent for sheaves on compact Hausdorff spaces
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。
These notes explain some descent results for $\infty$-categories of sheaves on compact Hausdorff spaces and derive some consequences. Specifically, given a compactly assembled $\infty$-category $\mathcal{E}$, we show that the functor sending a locally compact Hausdorff space $X$ to the $\infty$-category $\operatorname{Sh}^{\operatorname{post}}(X;\mathcal{E})$ of Postnikov complete $\mathcal{E}$-valued sheaves on $ X $ satisfies descent for proper surjections. This implies proper descent for left complete derived $\infty$-categories and that the functor $\operatorname{Sh}^{\operatorname{post}}(-;\mathcal{E})$ is a sheaf on the category of compact Hausdorff spaces equipped with the topology of finite jointly surjective families. Using this, we explain how to embed Postnikov complete sheaves on a locally compact Hausdorff space into condensed objects. This implies that the condensed and sheaf cohomologies of a locally compact Hausdorff space agree.