论文标题

部分可观测时空混沌系统的无模型预测

Heat kernel asymptotics for real powers of Laplacians

论文作者

Anghel, Cipriana

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 describe the small-time heat kernel asymptotics of real powers $Δ^r$, $r \in (0,1)$ of a non-negative self-adjoint generalized Laplacian $Δ$ acting on the sections of a hermitian vector bundle $\mathcal E$ over a closed oriented manifold $M$. First we treat separately the asymptotic on the diagonal of $M \times M$ and in a compact set away from it. Logarithmic terms appear only if $n$ is odd and $r$ is rational with even denominator. We prove the non-triviality of the coefficients appearing in the diagonal asymptotics, and also the non-locality of some of the coefficients. In the special case $r=1/2$, we give a simultaneous formula by proving that the heat kernel of $Δ^{1/2}$ is a polyhomogeneous conormal section in $\mathcal E \boxtimes {\mathcal E}^* $ on the standard blow-up space $M_{heat}$ of the diagonal at time $t=0$ inside $[0,\infty)\times M \times M$.

扫码加入交流群

加入微信交流群

微信交流群二维码

扫码加入学术交流群,获取更多资源