论文标题
Banach价值Sobolev函数的扩展和近似值
Extensions and approximations of Banach-valued Sobolev functions
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。
In complete metric measure spaces equipped with a doubling measure and supporting a weak Poincaré inequality, we investigate when a given Banach-valued Sobolev function defined on a subset satisfying a measure-density condition is the restriction of a Banach-valued Sobolev function defined on the whole space. We investigate the problem for Hajłasz- and Newton-Sobolev spaces, respectively. First, we show that Hajłasz-Sobolev extendability is independent of the target Banach spaces. We also show that every $c_0$-valued Newton-Sobolev extension set is a Banach-valued Newton-Sobolev extension set for every Banach space. We also prove that any measurable set satisfying a measure-density condition and a weak Poincaré inequality up to some scale is a Banach-valued Newton-Sobolev extension set for every Banach space. Conversely, we verify a folklore result stating that when $n\leq p<\infty$, every $W^{1,p}$-extension domain $Ω\subset \mathbb{R}^n$ supports a weak $(1,p)$-Poincaré inequality up to some scale. As a related result of independent interest, we prove that in any metric measure space when $1 \leq p < \infty$ and real-valued Lipschitz functions with bounded support are norm-dense in the real-valued $W^{1,p}$-space, then Banach-valued Lipschitz functions with bounded support are energy-dense in every Banach-valued $W^{1,p}$-space whenever the Banach space has the so-called metric approximation property.