Matches in SemOpenAlex for { <https://semopenalex.org/work/W4297813730> ?p ?o ?g. }
Showing items 1 to 83 of
83
with 100 items per page.
- W4297813730 abstract "Deep unfolding is a promising deep-learning technique, whose network architecture is based on expanding the recursive structure of existing iterative algorithms. Although convergence acceleration is a remarkable advantage of deep unfolding, its theoretical aspects have not been revealed yet. The first half of this study details the theoretical analysis of the convergence acceleration in deep-unfolded gradient descent (DUGD) whose trainable parameters are step sizes. We propose a plausible interpretation of the learned step-size parameters in DUGD by introducing the principle of Chebyshev steps derived from Chebyshev polynomials. The use of Chebyshev steps in gradient descent (GD) enables us to bound the spectral radius of a matrix governing the convergence speed of GD, leading to a tight upper bound on the convergence rate. The convergence rate of GD using Chebyshev steps is shown to be asymptotically optimal, although it has no momentum terms. We also show that Chebyshev steps numerically explain the learned step-size parameters in DUGD well. In the second half of the study, %we apply the theory of Chebyshev steps and Chebyshev-periodical successive over-relaxation (Chebyshev-PSOR) is proposed for accelerating linear/nonlinear fixed-point iterations. Theoretical analysis and numerical experiments indicate that Chebyshev-PSOR exhibits significantly faster convergence for various examples such as Jacobi method and proximal gradient methods." @default.
- W4297813730 created "2022-10-01" @default.
- W4297813730 creator A5049268989 @default.
- W4297813730 creator A5064555693 @default.
- W4297813730 date "2020-10-26" @default.
- W4297813730 modified "2023-09-26" @default.
- W4297813730 title "Convergence Acceleration via Chebyshev Step: Plausible Interpretation of Deep-Unfolded Gradient Descent" @default.
- W4297813730 doi "https://doi.org/10.48550/arxiv.2010.13335" @default.
- W4297813730 hasPublicationYear "2020" @default.
- W4297813730 type Work @default.
- W4297813730 citedByCount "0" @default.
- W4297813730 crossrefType "posted-content" @default.
- W4297813730 hasAuthorship W4297813730A5049268989 @default.
- W4297813730 hasAuthorship W4297813730A5064555693 @default.
- W4297813730 hasBestOaLocation W42978137301 @default.
- W4297813730 hasConcept C10628310 @default.
- W4297813730 hasConcept C117896860 @default.
- W4297813730 hasConcept C121332964 @default.
- W4297813730 hasConcept C122107654 @default.
- W4297813730 hasConcept C126255220 @default.
- W4297813730 hasConcept C127162648 @default.
- W4297813730 hasConcept C129785596 @default.
- W4297813730 hasConcept C134306372 @default.
- W4297813730 hasConcept C142484921 @default.
- W4297813730 hasConcept C145242015 @default.
- W4297813730 hasConcept C153258448 @default.
- W4297813730 hasConcept C154945302 @default.
- W4297813730 hasConcept C162324750 @default.
- W4297813730 hasConcept C171326582 @default.
- W4297813730 hasConcept C196512905 @default.
- W4297813730 hasConcept C21424316 @default.
- W4297813730 hasConcept C21736991 @default.
- W4297813730 hasConcept C2777303404 @default.
- W4297813730 hasConcept C28826006 @default.
- W4297813730 hasConcept C31258907 @default.
- W4297813730 hasConcept C33923547 @default.
- W4297813730 hasConcept C41008148 @default.
- W4297813730 hasConcept C50522688 @default.
- W4297813730 hasConcept C50644808 @default.
- W4297813730 hasConcept C57869625 @default.
- W4297813730 hasConcept C74650414 @default.
- W4297813730 hasConceptScore W4297813730C10628310 @default.
- W4297813730 hasConceptScore W4297813730C117896860 @default.
- W4297813730 hasConceptScore W4297813730C121332964 @default.
- W4297813730 hasConceptScore W4297813730C122107654 @default.
- W4297813730 hasConceptScore W4297813730C126255220 @default.
- W4297813730 hasConceptScore W4297813730C127162648 @default.
- W4297813730 hasConceptScore W4297813730C129785596 @default.
- W4297813730 hasConceptScore W4297813730C134306372 @default.
- W4297813730 hasConceptScore W4297813730C142484921 @default.
- W4297813730 hasConceptScore W4297813730C145242015 @default.
- W4297813730 hasConceptScore W4297813730C153258448 @default.
- W4297813730 hasConceptScore W4297813730C154945302 @default.
- W4297813730 hasConceptScore W4297813730C162324750 @default.
- W4297813730 hasConceptScore W4297813730C171326582 @default.
- W4297813730 hasConceptScore W4297813730C196512905 @default.
- W4297813730 hasConceptScore W4297813730C21424316 @default.
- W4297813730 hasConceptScore W4297813730C21736991 @default.
- W4297813730 hasConceptScore W4297813730C2777303404 @default.
- W4297813730 hasConceptScore W4297813730C28826006 @default.
- W4297813730 hasConceptScore W4297813730C31258907 @default.
- W4297813730 hasConceptScore W4297813730C33923547 @default.
- W4297813730 hasConceptScore W4297813730C41008148 @default.
- W4297813730 hasConceptScore W4297813730C50522688 @default.
- W4297813730 hasConceptScore W4297813730C50644808 @default.
- W4297813730 hasConceptScore W4297813730C57869625 @default.
- W4297813730 hasConceptScore W4297813730C74650414 @default.
- W4297813730 hasLocation W42978137301 @default.
- W4297813730 hasOpenAccess W4297813730 @default.
- W4297813730 hasPrimaryLocation W42978137301 @default.
- W4297813730 hasRelatedWork W1976132774 @default.
- W4297813730 hasRelatedWork W2021099221 @default.
- W4297813730 hasRelatedWork W2025186172 @default.
- W4297813730 hasRelatedWork W2035026539 @default.
- W4297813730 hasRelatedWork W2054912819 @default.
- W4297813730 hasRelatedWork W2270587440 @default.
- W4297813730 hasRelatedWork W3098848238 @default.
- W4297813730 hasRelatedWork W4297813730 @default.
- W4297813730 hasRelatedWork W49014452 @default.
- W4297813730 hasRelatedWork W88219822 @default.
- W4297813730 isParatext "false" @default.
- W4297813730 isRetracted "false" @default.
- W4297813730 workType "article" @default.