Representation and Characterization of Non-Stationary Processes by Dilation Operators and Induced Shape Space Manifolds
Ma\"el Dugast, Guillaume Bouleux, Eric Marcon

TL;DR
This paper introduces a geometric framework using dilation operators and shape space manifolds to analyze non-stationary stochastic processes, capturing their features through rotation matrices and curve analysis on manifolds.
Contribution
It develops a novel geometric approach to characterize non-stationary processes via dilation matrices and shape spaces, extending classical methods to non-stationary signals.
Findings
Dilation matrices encode process information regardless of stationarity.
The shape space framework allows comparison of process features through curves on manifolds.
The approach captures periodicity and other signal features in the geometry of dilation matrices.
Abstract
We have introduce a new vision of stochastic processes through the geometry induced by the dilation. The dilation matrices of a given processes are obtained by a composition of rotations matrices, contain the measure information in a condensed way. Particularly interesting is the fact that the obtention of dilation matrices is regardless of the stationarity of the underlying process. When the process is stationary, it coincides with the Naimark Dilation and only one rotation matrix is computed, when the process is non-stationary, a set of rotation matrices are computed. In particular, the periodicity of the correlation function that may appear in some classes of signal is transmitted to the set of dilation matrices. These rotation matrices, which can be arbitrarily close to each other depending on the sampling or the rescaling of the signal are seen as a distinctive feature of the…
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
