Karhunen-Lo\`eve Transforms of Autoregressive Windows Based on Matched Symmetry Groups
En palabras de los autores
The Karhunen-Lo\`eve transform of a window of a stationary autoregressive process of any order is described through its poles, extending the Algebraic Diversity framework, which selects a transform by the matched group. For simple roots every basis vector is a sum of at most sinusoids, at the frequencies where the spectrum equals the eigenvalue, plus modes decaying from the ends. The window covariance differs from a member of the class diagonalized by the type-II discrete cosine transform by a closed-form matrix of rank , exactly when no pole is zero, so the eigenvalues interlace spectrum samples on the cosine grid, and the coding-gain loss of the cosine transform is an exact function of that matrix with computable bounds: for the Markov block of image coding at pole it is below 1.2 percent. By Kronecker's theorem the rank is at most for rational spectra; a moving-average window departs in its Hankel corners, and an integrated (ARIMA) window departs by low rank from the type-VII sine class, exact for the random walk. The exact symmetry is generically the reflection and contains a wreath product for interleaved windows; the dihedral group of the even extension is matched approximately, the dominant pole's sign choosing cosine or sine. The two groups give the transform a fast form: a parity fold, half-length cosine cores and a cascade of at most first-order Cauchy stages, exactly for a minimal AR() model. The first-order case recovers results obtained by other means.
Apareció: martes, 10 de noviembre. arXiv. Preprint, todavía sin revisión por pares.
Comentario de los autores: v1: 38 pages, 7 figures