Complete set of translation invariant measurements with Lipschitz bounds

DOI10.1016/J.ACHA.2020.05.007zbMATH Open1448.94052arXiv1903.02811OpenAlexW3029118496MaRDI QIDQ778023FDOQ778023

Authors: Jameson Cahill, Andres Contreras, Andres A. Contreras Hip 

Publication date: 30 June 2020

Published in: Applied and Computational Harmonic Analysis (Search for Journal in Brave)

Abstract: In image and audio signal classification, a major problem is to build stable representations that are invariant under rigid motions and, more generally, to small diffeomorphisms. Translation invariant representations of signals in $mathbb{C}^n$ are of particular importance. The existence of such representations is intimately related to classical invariant theory, inverse problems in compressed sensing and deep learning. Despite an impressive body of litereature on the subject, most representations available are either: i) not stable due to the presence of high frequencies; ii) non discriminative; iii) non invariant when projected to finite dimensional subspaces. In the present paper, we construct low dimensional representations of signals in $mathbb{C}^n$ that are invariant under finite unitary group actions, as a special case we establish the existence of low-dimensional and complete $mathbb{Z}_m$-invariant representations for any $minmathbbN$. Our construction yields a stable, discriminative transform with semi-explicit Lipschitz bounds on the dimension; this is particularly relevant for applications. Using some tools from Algebraic Geometry, we define a high dimensional homogeneous function that is injective. We then exploit the projective character of this embedding and see that the target space can be reduced significantly by using a generic linear transformation. Finally, we introduce the notion of {it non-parallel} map, which is enjoyed by our function and employ this to construct a Lipschitz modification of it.

Full work available at URL: https://arxiv.org/abs/1903.02811

Recommendations

• Group invariant scattering
• Rotationally invariant time-frequency scattering transforms
• scientific article; zbMATH DE number 2090786
• Compactification of the rigid motions group in image processing
• Group invariance, stability to deformations, and complexity of deep convolutional representations

zbMATH Keywords

dimension reductioninvariant theorysignal classificationstable representation

Mathematics Subject Classification ID

Nontrigonometric harmonic analysis involving wavelets and other special systems (42C40) Signal theory (characterization, reconstruction, filtering, etc.) (94A12) General theory of group and pseudogroup actions (22F05) Linear operators in reproducing-kernel Hilbert spaces (including de Branges, de Branges-Rovnyak, and other structured spaces) (47B32)

Cites Work

• Group invariance, stability to deformations, and complexity of deep convolutional representations
• Learning deep architectures for AI
• Stable signal recovery from incomplete and inaccurate measurements
• A course in metric geometry
• Compressed sensing
• Group invariant scattering
• Saving phase: injectivity and stability for phase retrieval
• On signal reconstruction without phase
• An algebraic characterization of injectivity in phase retrieval
• Computational invariant theory. With two appendices by Vladimir L. Popov and an addendum by Nobert A. Campo and Vladimir L. Popov
• Generalized phase retrieval: measurement number, matrix recovery and beyond
• Separating invariants and finite reflection groups
• Multiscale sparse microcanonical models
• Separating invariants of finite groups
• Phase retrieval in infinite-dimensional Hilbert spaces
• Wavelet Scattering Regression of Quantum Chemical Energies
• On Lipschitz analysis and Lipschitz synthesis for the phase retrieval problem

Cited In (6)

• Towards a bilipschitz invariant theory
• Lie PCA: density estimation for symmetric manifolds
• Compactification of the rigid motions group in image processing
• Group invariant scattering
• Stable separation of orbits for finite abelian group actions
• Max filtering with reflection groups

Uses Software

• ckn_kernel

This page was built for publication: Complete set of translation invariant measurements with Lipschitz bounds

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q778023)