FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2317361501> ?p ?o ?g. }

• W2317361501 abstract "The goal of sparse coding is to compute efficient representations of signals. The problem arises in diverse fields, ranging from theoretical neuroscience to signal processing. Current approaches cast sparse coding as an optimization problem to which standard methods are applied. These approaches ignore the geometric structure that underlies many classes of signals. In this dissertation we develop a geometric approach to sparse coding that exploits this structure. Many classes of signal are observed to lie on low-dimensional manifolds in signal space. Capturing this structure is essential if sparse representations are to be computed. We show that if the dictionary, that is, the set of representational elements, is adapted to the signal manifold, and if minimum e1 -norm signal representations are computed, then it is possible to almost always compute sparse, that is, minimum e0-norm, signal representations and furthermore those representations will be as sparse as possible. To compute minimum e1-norm signal representations, we develop a new algorithm which we call Greedy Basis Pursuit (GBP). GBP is derived from a computational geometry and is equivalent to linear programming. We demonstrate that in some cases, GBP is capable of computing minimum e 1-norm signal representations faster than standard linear programming methods. Finally, to capture the geometric structure of the signal manifold we consider manifold learning approaches to dictionary learning. In particular, we experiment with clustering algorithms and show that the resulting dictionaries can be computed considerably faster than standard methods and that the results are often comparable." @default.
• W2317361501 created "2016-06-24" @default.
• W2317361501 creator A5012924253 @default.
• W2317361501 creator A5027843053 @default.
• W2317361501 date "2005-01-01" @default.
• W2317361501 modified "2023-09-24" @default.
• W2317361501 title "Sparse coding via geometry" @default.
• W2317361501 cites W109907426 @default.
• W2317361501 cites W1167735 @default.
• W2317361501 cites W125437831 @default.
• W2317361501 cites W140561259 @default.
• W2317361501 cites W141673549 @default.
• W2317361501 cites W1481866092 @default.
• W2317361501 cites W1486547850 @default.
• W2317361501 cites W1487193144 @default.
• W2317361501 cites W1491458120 @default.
• W2317361501 cites W1491710339 @default.
• W2317361501 cites W1494599013 @default.
• W2317361501 cites W1500551892 @default.
• W2317361501 cites W1506013575 @default.
• W2317361501 cites W1518078339 @default.
• W2317361501 cites W1521793179 @default.
• W2317361501 cites W1536929369 @default.
• W2317361501 cites W1548802052 @default.
• W2317361501 cites W1564605343 @default.
• W2317361501 cites W1568281949 @default.
• W2317361501 cites W1574845294 @default.
• W2317361501 cites W1575824550 @default.
• W2317361501 cites W1578196132 @default.
• W2317361501 cites W1578260139 @default.
• W2317361501 cites W1580157800 @default.
• W2317361501 cites W1586759373 @default.
• W2317361501 cites W1587863748 @default.
• W2317361501 cites W1597196612 @default.
• W2317361501 cites W1604668941 @default.
• W2317361501 cites W1607102876 @default.
• W2317361501 cites W1673584506 @default.
• W2317361501 cites W1677489248 @default.
• W2317361501 cites W1679913846 @default.
• W2317361501 cites W1757333299 @default.
• W2317361501 cites W1840528424 @default.
• W2317361501 cites W1902027874 @default.
• W2317361501 cites W1914401667 @default.
• W2317361501 cites W1963434311 @default.
• W2317361501 cites W1966240450 @default.
• W2317361501 cites W1968657806 @default.
• W2317361501 cites W1970554717 @default.
• W2317361501 cites W1971258090 @default.
• W2317361501 cites W1971507640 @default.
• W2317361501 cites W1973033359 @default.
• W2317361501 cites W1974049455 @default.
• W2317361501 cites W1975341768 @default.
• W2317361501 cites W1979191599 @default.
• W2317361501 cites W1980977871 @default.
• W2317361501 cites W1981315194 @default.
• W2317361501 cites W1982094581 @default.
• W2317361501 cites W1983939602 @default.
• W2317361501 cites W1984431842 @default.
• W2317361501 cites W1984931175 @default.
• W2317361501 cites W1988403812 @default.
• W2317361501 cites W1990905942 @default.
• W2317361501 cites W19916697 @default.
• W2317361501 cites W1992419399 @default.
• W2317361501 cites W1992902854 @default.
• W2317361501 cites W1993197592 @default.
• W2317361501 cites W1994154911 @default.
• W2317361501 cites W1995875735 @default.
• W2317361501 cites W1996021349 @default.
• W2317361501 cites W1996355918 @default.
• W2317361501 cites W1996494046 @default.
• W2317361501 cites W1996830635 @default.
• W2317361501 cites W1997146646 @default.
• W2317361501 cites W1997777356 @default.
• W2317361501 cites W2000358470 @default.
• W2317361501 cites W2000405398 @default.
• W2317361501 cites W2000420612 @default.
• W2317361501 cites W2000455396 @default.
• W2317361501 cites W2001141328 @default.
• W2317361501 cites W2003856348 @default.
• W2317361501 cites W2004775129 @default.
• W2317361501 cites W2006500012 @default.
• W2317361501 cites W2008772536 @default.
• W2317361501 cites W2010669260 @default.
• W2317361501 cites W2011039300 @default.
• W2317361501 cites W2011191879 @default.
• W2317361501 cites W2012762214 @default.
• W2317361501 cites W2013239224 @default.
• W2317361501 cites W2014698954 @default.
• W2317361501 cites W2014906057 @default.
• W2317361501 cites W2016479629 @default.
• W2317361501 cites W2017399882 @default.
• W2317361501 cites W2017726114 @default.
• W2317361501 cites W2017950741 @default.
• W2317361501 cites W2021302824 @default.
• W2317361501 cites W2023713450 @default.
• W2317361501 cites W2024223506 @default.
• W2317361501 cites W2024418895 @default.
• W2317361501 cites W2026559573 @default.
• W2317361501 cites W2028750032 @default.
• W2317361501 cites W2028781966 @default.

• next