Stochastic model reduction by aggregation

From MaRDI portal





[For the entire collection see Zbl 0745.00051.] In this note is given an example of a linear system evolving in a Hilbert space \(H\), \[ \begin{aligned} \dot z(t) & = Fz(t)+Gu(t),\quad z_ 0\in H\\ y(t) & = Hx(t)\in R^ m\\ \text{such that the projection \(x(t)\) of \(z(t)\) on a finite-dimensional subspace \(X\) of \(H\) satisfies} \dot x(t) & = Ax(t)+Bu(t)+Kv(t),\quad x_ 0\in X\\ \text{reproducing the same output through the equation} y(t) & = Cx(t)+Dv(t),\end{aligned} \] where \(v\) and \(x_ 0\) depend on \(z_ 0\), and conversely every solution of the latter system for some \(v\) and \(x_ 0\) is the projection on \(X\) of a solution of the former system starting from a suitable \(z_ 0\). The interpretation is that \(v\) acts as a noise term on the reduced system. The example is the wave equation on the positive half line with a boundary condition which is related to the solution of a finite- dimensional linear differential equation. The interesting feature is that the coefficients of the reduced system change non-trivially when the backward dynamics is considered. The idea is inspired by scattering theory techniques.











This page was built for publication: Stochastic model reduction by aggregation

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4026320)