Early stopping for statistical inverse problems via truncated SVD estimation

10/19/2017
by   Gilles Blanchard, et al.
0

We consider truncated SVD (or spectral cut-off, projection) estimators for a prototypical statistical inverse problem in dimension D. Since calculating the singular value decomposition (SVD) only for the largest singular values is much less costly than the full SVD, our aim is to select a data-driven truncation level m∈{1,...,D} only based on the knowledge of the first m singular values and vectors. We analyse in detail whether sequential early stopping rules of this type can preserve statistical optimality. Information-constrained lower bounds and matching upper bounds for a residual based stopping rule are provided, which give a clear picture in which situation optimal sequential adaptation is feasible. Finally, a hybrid two-step approach is proposed which allows for classical oracle inequalities while considerably reducing numerical complexity.

READ FULL TEXT

page 1

page 2

page 3

page 4

research
09/30/2019

Smoothed residual stopping for statistical inverse problems via truncated SVD estimation

This work examines under what circumstances adaptivity for truncated SVD...
research
09/12/2023

High Order Numerical Methods To Approximate The Singular Value Decomposition

In this paper, we present a class of high order methods to approximate t...
research
03/06/2018

Flip-Flop Spectrum-Revealing QR Factorization and Its Applications on Singular Value Decomposition

We present Flip-Flop Spectrum-Revealing QR (Flip-Flop SRQR) factorizatio...
research
10/13/2020

Projection techniques to update the truncated SVD of evolving matrices

This paper considers the problem of updating the rank-k truncated Singul...
research
02/25/2022

A Probabilistic Oracle Inequality and Quantification of Uncertainty of a modified Discrepancy Principle for Statistical Inverse Problems

In this note we consider spectral cut-off estimators to solve a statisti...
research
03/28/2023

SVD-DIP: Overcoming the Overfitting Problem in DIP-based CT Reconstruction

The deep image prior (DIP) is a well-established unsupervised deep learn...
research
08/13/2021

A Parallel Distributed Algorithm for the Power SVD Method

In this work, we study how to implement a distributed algorithm for the ...

Please sign up or login with your details

Forgot password? Click here to reset