论文标题
粗糙路径微分方程产生的流量的几何分解
Geometric decomposition of flows generated by rough path differential equations
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。
Whenever an Itô-Wentsel type of formula holds for composition of flows of a certain differential dynamics, there exists locally a decomposition of the corresponding flow according to complementary distributions (or foliations, in the case of integrability of these distributions). Many examples have been proved in distinct context of dynamics: Stratonovich stochastic equations, Lévy driven noise, low regularity $α$-Hölder control functions ($ α\in (1/2,1]$), see e.g. [6], [7], [20], [21]. Here we present the proof of this categorical property: we illustrate with the $α$-Hölder rough path, $α\in (1/3, 1/2]$ using the Itô-Wentsel formula in this context proved in [5]. Different from the previous approaches, here however, instead of using an intrinsic rough path calculus on manifolds, the manifold has to be embedded in an Euclidean space. A cascade decomposition is also shown when we have multiple lower dimensional directions which span the whole space. As application, the linear case is treated in details: the cascade decomposition provides a row factorization of all matrices which allow real logarithm.
