Equivariant Deep Equilibrium Models for Imaging Inverse Problems

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Mehta, Alexander, Kitichotkul, Ruangrawee, Goyal, Vivek K, Tachella, Julián
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914168544690176
author Mehta, Alexander
Kitichotkul, Ruangrawee
Goyal, Vivek K
Tachella, Julián
author_facet Mehta, Alexander
Kitichotkul, Ruangrawee
Goyal, Vivek K
Tachella, Julián
contents Equivariant imaging (EI) enables training signal reconstruction models without requiring ground truth data by leveraging signal symmetries. Deep equilibrium models (DEQs) are a powerful class of neural networks where the output is a fixed point of a learned operator. However, training DEQs with complex EI losses requires implicit differentiation through fixed-point computations, whose implementation can be challenging. We show that backpropagation can be implemented modularly, simplifying training. Experiments demonstrate that DEQs trained with implicit differentiation outperform those trained with Jacobian-free backpropagation and other baseline methods. Additionally, we find evidence that EI-trained DEQs approximate the proximal map of an invariant prior.
format Preprint
id arxiv_https___arxiv_org_abs_2511_18667
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Equivariant Deep Equilibrium Models for Imaging Inverse Problems
Mehta, Alexander
Kitichotkul, Ruangrawee
Goyal, Vivek K
Tachella, Julián
Image and Video Processing
Machine Learning
Signal Processing
Equivariant imaging (EI) enables training signal reconstruction models without requiring ground truth data by leveraging signal symmetries. Deep equilibrium models (DEQs) are a powerful class of neural networks where the output is a fixed point of a learned operator. However, training DEQs with complex EI losses requires implicit differentiation through fixed-point computations, whose implementation can be challenging. We show that backpropagation can be implemented modularly, simplifying training. Experiments demonstrate that DEQs trained with implicit differentiation outperform those trained with Jacobian-free backpropagation and other baseline methods. Additionally, we find evidence that EI-trained DEQs approximate the proximal map of an invariant prior.
title Equivariant Deep Equilibrium Models for Imaging Inverse Problems
topic Image and Video Processing
Machine Learning
Signal Processing
url https://arxiv.org/abs/2511.18667