FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W4319459525> ?p ?o ?g. }

• W4319459525 endingPage "952" @default.
• W4319459525 startingPage "897" @default.
• W4319459525 abstract "Non-smooth optimization is a core ingredient of many imaging or machine learning pipelines. Non-smoothness encodes structural constraints on the solutions, such as sparsity, group sparsity, low-rank and sharp edges. It is also the basis for the definition of robust loss functions and scale-free functionals such as square-root Lasso. Standard approaches to deal with non-smoothness leverage either proximal splitting or coordinate descent. These approaches are effective but usually require parameter tuning, preconditioning or some sort of support pruning. In this work, we advocate and study a different route, which operates a non-convex but smooth over-parameterization of the underlying non-smooth optimization problems. This generalizes quadratic variational forms that are at the heart of the popular Iterative Reweighted Least Squares. Our main theoretical contribution connects gradient descent on this reformulation to a mirror descent flow with a varying Hessian metric. This analysis is crucial to derive convergence bounds that are dimension-free. This explains the efficiency of the method when using small grid sizes in imaging. Our main algorithmic contribution is to apply the Variable Projection method which defines a new formulation by explicitly minimizing over part of the variables. This leads to a better conditioning of the minimized functional and improves the convergence of simple but very efficient gradient-based methods, for instance quasi-Newton solvers. We exemplify the use of this new solver for the resolution of regularized regression problems for inverse problems and supervised learning, including total variation prior and non-convex regularizers." @default.
• W4319459525 created "2023-02-09" @default.
• W4319459525 creator A5004350216 @default.
• W4319459525 creator A5058651667 @default.
• W4319459525 date "2023-02-08" @default.
• W4319459525 modified "2023-09-25" @default.
• W4319459525 title "Smooth over-parameterized solvers for non-smooth structured optimization" @default.
• W4319459525 cites W1976618413 @default.
• W4319459525 cites W1983677661 @default.
• W4319459525 cites W2003372231 @default.
• W4319459525 cites W2016384870 @default.
• W4319459525 cites W2018089423 @default.
• W4319459525 cites W2019569173 @default.
• W4319459525 cites W2033603511 @default.
• W4319459525 cites W2039939700 @default.
• W4319459525 cites W2040763577 @default.
• W4319459525 cites W2045704273 @default.
• W4319459525 cites W2047003499 @default.
• W4319459525 cites W2058737544 @default.
• W4319459525 cites W2065180801 @default.
• W4319459525 cites W2072081687 @default.
• W4319459525 cites W2076605490 @default.
• W4319459525 cites W2080744942 @default.
• W4319459525 cites W2088616581 @default.
• W4319459525 cites W2090636411 @default.
• W4319459525 cites W2091918879 @default.
• W4319459525 cites W2092663520 @default.
• W4319459525 cites W2093545205 @default.
• W4319459525 cites W2100556411 @default.
• W4319459525 cites W2101868363 @default.
• W4319459525 cites W2103559027 @default.
• W4319459525 cites W2115706991 @default.
• W4319459525 cites W2118550318 @default.
• W4319459525 cites W2119883478 @default.
• W4319459525 cites W2127447444 @default.
• W4319459525 cites W2128610923 @default.
• W4319459525 cites W2138019504 @default.
• W4319459525 cites W2145096794 @default.
• W4319459525 cites W2145962650 @default.
• W4319459525 cites W2155447220 @default.
• W4319459525 cites W2159494203 @default.
• W4319459525 cites W2160450758 @default.
• W4319459525 cites W2167400582 @default.
• W4319459525 cites W2168745297 @default.
• W4319459525 cites W2416562393 @default.
• W4319459525 cites W2546982352 @default.
• W4319459525 cites W2551153860 @default.
• W4319459525 cites W2585868105 @default.
• W4319459525 cites W2607685496 @default.
• W4319459525 cites W2616577182 @default.
• W4319459525 cites W2895545984 @default.
• W4319459525 cites W2964137109 @default.
• W4319459525 cites W2964189626 @default.
• W4319459525 cites W2964325628 @default.
• W4319459525 cites W2978118767 @default.
• W4319459525 cites W3090192109 @default.
• W4319459525 cites W3121832289 @default.
• W4319459525 cites W3184538210 @default.
• W4319459525 cites W3214386716 @default.
• W4319459525 cites W4213310170 @default.
• W4319459525 cites W4254546220 @default.
• W4319459525 cites W4294541781 @default.
• W4319459525 doi "https://doi.org/10.1007/s10107-022-01923-3" @default.
• W4319459525 hasPublicationYear "2023" @default.
• W4319459525 type Work @default.
• W4319459525 citedByCount "0" @default.
• W4319459525 crossrefType "journal-article" @default.
• W4319459525 hasAuthorship W4319459525A5004350216 @default.
• W4319459525 hasAuthorship W4319459525A5058651667 @default.
• W4319459525 hasBestOaLocation W43194595252 @default.
• W4319459525 hasConcept C102634674 @default.
• W4319459525 hasConcept C10494615 @default.
• W4319459525 hasConcept C112680207 @default.
• W4319459525 hasConcept C11413529 @default.
• W4319459525 hasConcept C126255220 @default.
• W4319459525 hasConcept C134306372 @default.
• W4319459525 hasConcept C137836250 @default.
• W4319459525 hasConcept C153258448 @default.
• W4319459525 hasConcept C154945302 @default.
• W4319459525 hasConcept C157553263 @default.
• W4319459525 hasConcept C157972887 @default.
• W4319459525 hasConcept C203616005 @default.
• W4319459525 hasConcept C2524010 @default.
• W4319459525 hasConcept C2778770139 @default.
• W4319459525 hasConcept C28826006 @default.
• W4319459525 hasConcept C33923547 @default.
• W4319459525 hasConcept C41008148 @default.
• W4319459525 hasConcept C50644808 @default.
• W4319459525 hasConceptScore W4319459525C102634674 @default.
• W4319459525 hasConceptScore W4319459525C10494615 @default.
• W4319459525 hasConceptScore W4319459525C112680207 @default.
• W4319459525 hasConceptScore W4319459525C11413529 @default.
• W4319459525 hasConceptScore W4319459525C126255220 @default.
• W4319459525 hasConceptScore W4319459525C134306372 @default.
• W4319459525 hasConceptScore W4319459525C137836250 @default.
• W4319459525 hasConceptScore W4319459525C153258448 @default.
• W4319459525 hasConceptScore W4319459525C154945302 @default.
• W4319459525 hasConceptScore W4319459525C157553263 @default.

• next