Learning linearized assignment flows for image labeling

DOI10.1007/S10851-022-01132-9arXiv2108.02571OpenAlexW3189781688MaRDI QIDQ6156048FDOQ6156048

Authors: Alexander Zeilmann, Stefania Petra, Christoph Schnörr 

Publication date: 12 June 2023

Published in: Journal of Mathematical Imaging and Vision (Search for Journal in Brave)

Abstract: We introduce a novel algorithm for estimating optimal parameters of linearized assignment flows for image labeling. An exact formula is derived for the parameter gradient of any loss function that is constrained by the linear system of ODEs determining the linearized assignment flow. We show how to efficiently evaluate this formula using a Krylov subspace and a low-rank approximation. This enables us to perform parameter learning by Riemannian gradient descent in the parameter space, without the need to backpropagate errors or to solve an adjoint equation. Experiments demonstrate that our method performs as good as highly-tuned machine learning software using automatic differentiation. Unlike methods employing automatic differentiation, our approach yields a low-dimensional representation of internal parameters and their dynamics which helps to understand how assignment flows and more generally neural networks work and perform.

Full work available at URL: https://arxiv.org/abs/2108.02571

zbMATH Keywords

low-rank approximationparameter learningassignment flowsexponential integrationimage labeling

Mathematics Subject Classification ID

Learning and adaptive systems in artificial intelligence (68T05) Image analysis in multivariate analysis (62H35) Computing methodologies for image processing (68U10) Evolutionary games (91A22) Ordinary differential equations and systems on manifolds (34C40) Computer science (68-XX) Information and communication theory, circuits (94-XX)

Cites Work

• Algorithm 919: A Krylov subspace algorithm for evaluating the \(\phi\)-functions appearing in exponential integrators
• Functions of Matrices
• Nineteen Dubious Ways to Compute the Exponential of a Matrix, Twenty-Five Years Later
• The ubiquitous Kronecker product
• Numerical methods for large eigenvalue problems
• Title not available (Why is that?)
• Ordinary differential equations and dynamical systems
• Title not available (Why is that?)
• Title not available (Why is that?)
• Exponential integrators
• Computing the action of the matrix exponential, with an application to exponential integrators
• Title not available (Why is that?)
• On Krylov Subspace Approximations to the Matrix Exponential Operator
• Analysis of Some Krylov Subspace Approximations to the Matrix Exponential Operator
• Exponential Rosenbrock-Type Methods
• Assignment flows
• Approximation of functions of large matrices with Kronecker structure
• Image labeling by assignment
• Title not available (Why is that?)
• Geometric numerical integration of the assignment flow
• Learning linear assignment flows for image labeling via exponential integration
• Computing low‐rank approximations of the Fréchet derivative of a matrix function using Krylov subspace methods
• Assignment flows for data labeling on graphs: convergence and stability

Cited In (3)

• Learning adaptive regularization for image labeling using geometric assignment
• A Nonlocal Graph-PDE and Higher-Order Geometric Integration for Image Labeling
• Unsupervised assignment flow: label learning on feature manifolds by spatially regularized geometric assignment

This page was built for publication: Learning linearized assignment flows for image labeling

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6156048)