Saved in:
Bibliographic Details
Main Authors: Buskulic, Nathan, Fadili, Jalal, Quéau, Yvain
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2403.05395
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917608365752320
author Buskulic, Nathan
Fadili, Jalal
Quéau, Yvain
author_facet Buskulic, Nathan
Fadili, Jalal
Quéau, Yvain
contents Advanced machine learning methods, and more prominently neural networks, have become standard to solve inverse problems over the last years. However, the theoretical recovery guarantees of such methods are still scarce and difficult to achieve. Only recently did unsupervised methods such as Deep Image Prior (DIP) get equipped with convergence and recovery guarantees for generic loss functions when trained through gradient flow with an appropriate initialization. In this paper, we extend these results by proving that these guarantees hold true when using gradient descent with an appropriately chosen step-size/learning rate. We also show that the discretization only affects the overparametrization bound for a two-layer DIP network by a constant and thus that the different guarantees found for the gradient flow will hold for gradient descent.
format Preprint
id arxiv_https___arxiv_org_abs_2403_05395
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Recovery Guarantees of Unsupervised Neural Networks for Inverse Problems trained with Gradient Descent
Buskulic, Nathan
Fadili, Jalal
Quéau, Yvain
Machine Learning
Advanced machine learning methods, and more prominently neural networks, have become standard to solve inverse problems over the last years. However, the theoretical recovery guarantees of such methods are still scarce and difficult to achieve. Only recently did unsupervised methods such as Deep Image Prior (DIP) get equipped with convergence and recovery guarantees for generic loss functions when trained through gradient flow with an appropriate initialization. In this paper, we extend these results by proving that these guarantees hold true when using gradient descent with an appropriately chosen step-size/learning rate. We also show that the discretization only affects the overparametrization bound for a two-layer DIP network by a constant and thus that the different guarantees found for the gradient flow will hold for gradient descent.
title Recovery Guarantees of Unsupervised Neural Networks for Inverse Problems trained with Gradient Descent
topic Machine Learning
url https://arxiv.org/abs/2403.05395