论文标题
部分可观测时空混沌系统的无模型预测
On rank 3 quadratic equations of projective varieties
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。
Let $X \subset ¶^r$ be a linearly normal variety defined by a very ample line bundle $L$ on a projective variety $X$. Recently it is shown in \cite{HLMP} that there are many cases where $(X,L)$ satisfies property $\textsf{QR} (3)$ in the sense that the homogeneous ideal $I(X,L)$ of $X$ is generated by quadratic polynomials of rank $3$. The locus $Φ_3 (X,L)$ of rank $3$ quadratic equations of $X$ in $¶\left( I(X,L)_2 \right)$ is a projective algebraic set, and property $\textsf{QR} (3)$ of $(X,L)$ is equivalent to that $Φ_3 (X)$ is nondegenerate in $¶\left( I(X)_2 \right)$. In this paper we study geometric structures of $Φ_3 (X,L)$ such as its minimal irreducible decomposition. Let \begin{equation*} Σ(X,L) = \{ (A,B) ~|~ A,B \in {\rm Pic}(X),~L = A^2 \otimes B,~h^0 (X,A) \geq 2,~h^0 (X,B) \geq 1 \}. \end{equation*} We first construct a projective subvariety $W(A,B) \subset Φ_3 (X,L)$ for each $(A,B)$ in $Σ(X,L)$. Then we prove that the equality \begin{equation*} Φ_3 (X,L) ~=~ \bigcup_{(A,B) \in Σ(X,L)} W(A,B) \end{equation*} holds when $X$ is locally factorial. Thus this is an irreducible decomposition of $Φ_3 (X,L)$ when ${\rm Pic} (X)$ is finitely generated and hence $Σ(X,L)$ is a finite set. Also we find a condition that the above irreducible decomposition is minimal. For example, it is a minimal irreducible decomposition of $Φ_3 (X,L)$ if ${\rm Pic}(X)$ is generated by a very ample line bundle.