-
On the convergence analysis of one-shot inversion methods
Authors:
Marcella Bonazzoli,
Houssem Haddar,
Tuan Anh Vu
Abstract:
When an inverse problem is solved by a gradient-based optimization algorithm, the corresponding forward and adjoint problems, which are introduced to compute the gradient, can be also solved iteratively. The idea of iterating at the same time on the inverse problem unknown and on the forward and adjoint problem solutions yields the concept of one-shot inversion methods. We are especially intereste…
▽ More
When an inverse problem is solved by a gradient-based optimization algorithm, the corresponding forward and adjoint problems, which are introduced to compute the gradient, can be also solved iteratively. The idea of iterating at the same time on the inverse problem unknown and on the forward and adjoint problem solutions yields the concept of one-shot inversion methods. We are especially interested in the case where the inner iterations for the direct and adjoint problems are incomplete, that is, stopped before achieving a high accuracy on their solutions. Here, we focus on general linear inverse problems and generic fixed-point iterations for the associated forward problem. We analyze variants of the so-called multi-step one-shot methods, in particular semi-implicit schemes with a regularization parameter. We establish sufficient conditions on the descent step for convergence, by studying the eigenvalues of the block matrix of the coupled iterations. Several numerical experiments are provided to illustrate the convergence of these methods in comparison with the classical gradient descent, where the forward and adjoint problems are solved exactly by a direct solver instead. We observe that very few inner iterations are enough to guarantee good convergence of the inversion algorithm, even in the presence of noisy data.
△ Less
Submitted 11 April, 2024;
originally announced April 2024.
-
Bounds on Functionality and Symmetric Difference -- Two Intriguing Graph Parameters
Authors:
Pavel Dvořák,
Lukáš Folwarczný,
Michal Opler,
Pavel Pudlák,
Robert Šámal,
Tung Anh Vu
Abstract:
[Alecu et al.: Graph functionality, JCTB2021] define functionality, a graph parameter that generalizes graph degeneracy. They research the relation of functionality to many other graph parameters (tree-width, clique-width, VC-dimension, etc.). Extending their research, we prove logarithmic lower bound for functionality of random graph $G(n,p)$ for large range of $p$. Previously known graphs have f…
▽ More
[Alecu et al.: Graph functionality, JCTB2021] define functionality, a graph parameter that generalizes graph degeneracy. They research the relation of functionality to many other graph parameters (tree-width, clique-width, VC-dimension, etc.). Extending their research, we prove logarithmic lower bound for functionality of random graph $G(n,p)$ for large range of $p$. Previously known graphs have functionality logarithmic in number of vertices. We show that for every graph $G$ on $n$ vertices we have $\mathrm{fun}(G) \le O(\sqrt{ n \log n})$ and we give a nearly matching $Ω(\sqrt{n})$-lower bound provided by projective planes.
Further, we study a related graph parameter \emph{symmetric difference}, the minimum of $|N(u) ΔN(v)|$ over all pairs of vertices of the ``worst possible'' induced subgraph. It was observed by Alecu et al. that $\mathrm{fun}(G) \le \mathrm{sd}(G)+1$ for every graph $G$. We compare $\mathrm{fun}$ and $\mathrm{sd}$ for the class $\mathrm{INT}$ of interval graphs and $\mathrm{CA}$ of circular-arc graphs. We let $\mathrm{INT}_n$ denote the $n$-vertex interval graphs, similarly for $\mathrm{CA}_n$. Alecu et al. ask, whether $\mathrm{fun}(\mathrm{INT})$ is bounded. Dallard et al. answer this positively in a recent preprint. On the other hand, we show that $Ω(\sqrt[4]{n}) \leq \mathrm{sd}(\mathrm{INT}_n) \leq O(\sqrt[3]{n})$. For the related class $\mathrm{CA}$ we show that $\mathrm{sd}(\mathrm{CA}_n) = Θ(\sqrt{n})$. We propose a follow-up question: is $\mathrm{fun}(\mathrm{CA})$ bounded?
△ Less
Submitted 23 February, 2023;
originally announced February 2023.
-
Convergence analysis of multi-step one-shot methods for linear inverse problems
Authors:
Marcella Bonazzoli,
Houssem Haddar,
Tuan Anh Vu
Abstract:
In this work we are interested in general linear inverse problems where the corresponding forward problem is solved iteratively using fixed point methods. Then one-shot methods, which iterate at the same time on the forward problem solution and on the inverse problem unknown, can be applied. We analyze two variants of the so-called multi-step one-shot methods and establish sufficient conditions on…
▽ More
In this work we are interested in general linear inverse problems where the corresponding forward problem is solved iteratively using fixed point methods. Then one-shot methods, which iterate at the same time on the forward problem solution and on the inverse problem unknown, can be applied. We analyze two variants of the so-called multi-step one-shot methods and establish sufficient conditions on the descent step for their convergence, by studying the eigenvalues of the block matrix of the coupled iterations. Several numerical experiments are provided to illustrate the convergence of these methods in comparison with the classical usual and shifted gradient descent. In particular, we observe that very few inner iterations on the forward problem are enough to guarantee good convergence of the inversion algorithm.
△ Less
Submitted 26 July, 2022; v1 submitted 21 July, 2022;
originally announced July 2022.
