Another key open problem is to derive explicit estimates for the difference between the optimal values of problems (2) and (1) as a function of the number N of data points,…
Statement
Another key open problem is to derive explicit estimates for the difference between the optimal values of problems (2) and (1) as a function of the number N of data points, the degree m of the approximating dictionary Ψ, and the data sampling step \tau . We wonder if recent advances in the quantitative error analysis for EDMD can allow for progress in this direction. This observation suggests that one should apply EDMD to approximate the Koop- man operator from the given data and replace the exact Lie derivative f \cdot ∇v in the optimization problem (1) with a data-driven approximation. Specifically, let b(x) and Ψ(x) be basis vectors for the spaces of polynomials with indeterminate x and degree d ad m \geq d, respectively. First, we approximate K\tau b \approx KmN \tau Ψ \tau where the matrix KmN solves the least-squares problem N 1 X min kb(yn ) − KΨ(xn )k2 . K N n=1 Then, we approximate the Lie derivative of any polynomial v(x) = c \cdot b(x) as \tau KmN Ψ − b f \cdot ∇v \approx \cdot c =: L\taumN v. \tau Finally, we compute approximate upper bounds on the maximal time average of an observable g by replacing the minimization problem in (1) with (2) inf U s.t. U − g(x) − L\taumN v(x) \geq 0 \forallx \in X. U, v
Record
Comments
Nothing has been published against this problem yet, and nothing has been said about it.
Sign in with an institutional address to take part in the discussion. Reading every thread stays open to everyone.
Sign inSolve with an agent
Open the statement in a chat, with the problem and the ground rules already written into the prompt.