Model Consistency of Partly Smooth Regularizers
From MaRDI portal
Abstract: This paper studies least-square regression penalized with partly smooth convex regularizers. This class of functions is very large and versatile allowing to promote solutions conforming to some notion of low-complexity. Indeed, they force solutions of variational problems to belong to a low-dimensional manifold (the so-called model) which is stable under small perturbations of the function. This property is crucial to make the underlying low-complexity model robust to small noise. We show that a generalized "irrepresentable condition" implies stable model selection under small noise perturbations in the observations and the design matrix, when the regularization parameter is tuned proportionally to the noise level. This condition is shown to be almost a necessary condition. We then show that this condition implies model consistency of the regularized estimator. That is, with a probability tending to one as the number of measurements increases, the regularized estimator belongs to the correct low-dimensional model manifold. This work unifies and generalizes several previous ones, where model consistency is known to hold for sparse, group sparse, total variation and low-rank regularizations.
Cited in
(26)- Monotone operator theory in convex optimization
- Exact matrix completion based on low rank Hankel structure in the Fourier domain
- Isotonicity of the proximity operator and mixed variational inequalities in Hilbert spaces
- On the interplay between acceleration and identification for the proximal gradient algorithm
- Block-based refitting in \(\ell_{12}\) sparse regularization
- Sharp oracle inequalities for low-complexity priors
- Local convergence properties of Douglas-Rachford and alternating direction method of multipliers
- The degrees of freedom of partly smooth regularizers
- Convergence of the forward-backward algorithm: beyond the worst-case with the help of geometry
- Low complexity regularization of linear inverse problems
- Activity identification and local linear convergence of forward-backward-type methods
- Spectral decompositions using one-homogeneous functionals
- Sensitivity analysis for mirror-stratifiable convex functions
- Active-set Newton methods and partial smoothness
- Isotonicity of the proximity operator and stochastic optimization problems in Hilbert quasi-lattices endowed with Lorentz cones
- Dual extrapolation for sparse GLMs
- Local linear convergence of proximal coordinate descent algorithm
- The geometry of monotone operator splitting methods
- Exact recovery of the support of piecewise constant images via total variation regularization
- Pattern recovery by SLOPE
- A unified framework for pattern recovery in penalized and thresholded estimation and its geometry
- Super-resolved Lasso
- Unveiling low-dimensional patterns induced by convex non-differentiable regularizers
- On reducibility of ADMM based on degenerate proximal point analysis
- A normal map-based proximal stochastic gradient method: convergence and identification properties
- Isotonicity of proximity operators in general quasi-lattices and optimization problems
This page was built for publication: Model Consistency of Partly Smooth Regularizers
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4566738)
