An Adaptively Inexact Method for Bilevel Learning Using Primal-Dual Style Differentiation

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Bogensperger, Lea, Ehrhardt, Matthias J., Pock, Thomas, Salehi, Mohammad Sadegh, Wong, Hok Shing
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866915332335075328
author Bogensperger, Lea
Ehrhardt, Matthias J.
Pock, Thomas
Salehi, Mohammad Sadegh
Wong, Hok Shing
author_facet Bogensperger, Lea
Ehrhardt, Matthias J.
Pock, Thomas
Salehi, Mohammad Sadegh
Wong, Hok Shing
contents We consider a bilevel learning framework for learning linear operators. In this framework, the learnable parameters are optimized via a loss function that also depends on the minimizer of a convex optimization problem (denoted lower-level problem). We utilize an iterative algorithm called `piggyback' to compute the gradient of the loss and minimizer of the lower-level problem. Given that the lower-level problem is solved numerically, the loss function and thus its gradient can only be computed inexactly. To estimate the accuracy of the computed hypergradient, we derive an a-posteriori error bound, which provides guides for setting the tolerance for the lower-level problem, as well as the piggyback algorithm. To efficiently solve the upper-level optimization, we also propose an adaptive method for choosing a suitable step-size. To illustrate the proposed method, we consider a few learned regularizer problems, such as training an input-convex neural network.
format Preprint
id arxiv_https___arxiv_org_abs_2412_06436
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle An Adaptively Inexact Method for Bilevel Learning Using Primal-Dual Style Differentiation
Bogensperger, Lea
Ehrhardt, Matthias J.
Pock, Thomas
Salehi, Mohammad Sadegh
Wong, Hok Shing
Optimization and Control
Machine Learning
We consider a bilevel learning framework for learning linear operators. In this framework, the learnable parameters are optimized via a loss function that also depends on the minimizer of a convex optimization problem (denoted lower-level problem). We utilize an iterative algorithm called `piggyback' to compute the gradient of the loss and minimizer of the lower-level problem. Given that the lower-level problem is solved numerically, the loss function and thus its gradient can only be computed inexactly. To estimate the accuracy of the computed hypergradient, we derive an a-posteriori error bound, which provides guides for setting the tolerance for the lower-level problem, as well as the piggyback algorithm. To efficiently solve the upper-level optimization, we also propose an adaptive method for choosing a suitable step-size. To illustrate the proposed method, we consider a few learned regularizer problems, such as training an input-convex neural network.
title An Adaptively Inexact Method for Bilevel Learning Using Primal-Dual Style Differentiation
topic Optimization and Control
Machine Learning
url https://arxiv.org/abs/2412.06436