Data-Based Reconstruction of Chaotic Systems by Stochastic Iterative Greedy Algorithm

Abstract EN

It is challenging to reconstruct a nonlinear dynamical system when sufficient observations are not available.

Recent study shows this problem can be solved by paradigm of compressive sensing.

In this paper, we study the reconstruction of chaotic systems based on the stochastic gradient matching pursuit (StoGradMP) method.

Comparing with the previous method based on convex optimization, the study results show that the StoGradMP method performs much better when the numerical sampling period is small.

So the present study enables potential application of the reconstruction method using limited observations in some special situations where limited observations can be acquired in limited time.