Regularization Theory of the Analytic Deep Prior Approach

Abstract

The analytic deep prior (ADP) approach was recently introduced for the theoretical analysis of deep image prior (DIP) methods with special network architectures. In this paper, we prove that ADP is in fact equivalent to classical variational Ivanov methods for solving ill-posed inverse problems. Besides, we propose a new variant which incorporates the strategy of early stopping into the ADP model. For both variants, we show how classical regularization properties (existence, stability, convergence) can be obtained under common assumptions.

References in corpus (2)

• Denoising and Regularization via Exploiting the Structural Bias of Convolutional Generators
• On Measuring and Controlling the Spectral Bias of the Deep Image Prior

Cited by in corpus (4)

• An Educated Warm Start For Deep Image Prior-Based Micro CT Reconstruction
• An adaptively inexact first-order method for bilevel optimization with application to hyperparameter learning
• Invertible residual networks in the context of regularization theory for linear inverse problems
• Fast Inexact Bilevel Optimization for Analytical Deep Image Priors