论文标题
部分可观测时空混沌系统的无模型预测
Interpolation Polynomials and Linear Algebra
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 reconsider the theory of Lagrange interpolation polynomials with multiple interpolation points and apply it to linear algebra. For instance, $A$ be a linear operator satisfying a degree $n$ polynomial equation $P(A)=0$. One can see that the evaluation of a meromorphic function $F$ at $A$ is equal to $Q(A)$, where $Q$ is the degree $<n$ interpolation polynomial of $F$ with the the set of interpolation points equal to the set of roots of the polynomial $P$. In particular, for $A$ an $n \times n$ matrix, there is a common belief that for computing $F(A)$ one has to reduce $A$ to its Jordan form. Let $P$ be the characteristic polynomial of $A$. Then by the Cayley-Hamilton theorem, $P(A)=0$. And thus the matrix $F(A)$ can be found without reducing $A$ to its Jordan form. Computation of the Jordan form for $A$ involves many extra computations. In the paper we show that it is not needed. One application is to compute the matrix exponential for a matrix with repeated eigenvalues, thereby solving arbitrary order linear differential equations with constant coefficients.
