Oracle inequalities for square root analysis estimators with application to total variation penalties
Preprint
Abstract
Through the direct study of the analysis estimator we derive oracle inequalities with fast and slow rates by adapting the arguments involving projections by Dalalyan, Hebiri and Lederer (2017). We then extend the theory to the square root analysis estimator. Finally, we focus on (square root) total variation regularized estimators on graphs and obtain constant-friendly rates, which, up to log-terms, match previous results obtained by entropy calculations. We also obtain an oracle inequality for the (square root) total variation regularized estimator over the cycle graph.Keywords
All Related Versions
This publication has 17 references indexed in Scilit:
- New Risk Bounds for 2D Total Variation DenoisingIEEE Transactions on Information Theory, 2021
- The Group Square-Root Lasso: Theoretical Properties and Fast AlgorithmsIEEE Transactions on Information Theory, 2013
- Robust Sparse Analysis RegularizationIEEE Transactions on Information Theory, 2012
- Scaled sparse linear regressionBiometrika, 2012
- Square-root lasso: pivotal recovery of sparse signals via conic programmingBiometrika, 2011
- The solution path of the generalized lassoThe Annals of Statistics, 2011
- Analysis versus synthesis in signal priorsInverse Problems, 2007
- Sparsity and Smoothness Via the Fused LassoJournal of the Royal Statistical Society Series B: Statistical Methodology, 2004
- Regression Shrinkage and Selection Via the LassoJournal of the Royal Statistical Society Series B: Statistical Methodology, 1996
- On the Generalized Inverse of an Incidence MatrixJournal of the Society for Industrial and Applied Mathematics, 1965