Matches in SemOpenAlex for { <https://semopenalex.org/work/W2892755733> ?p ?o ?g. }
Showing items 1 to 95 of
95
with 100 items per page.
- W2892755733 abstract "Randomly initialized first-order optimization algorithms are the method of choice for solving many high-dimensional nonconvex problems in machine learning, yet general theoretical guarantees cannot rule out convergence to critical points of poor objective value. For some highly structured nonconvex problems however, the success of gradient descent can be understood by studying the geometry of the objective. We study one such problem -- complete orthogonal dictionary learning, and provide converge guarantees for randomly initialized gradient descent to the neighborhood of a global optimum. The resulting rates scale as low order polynomials in the dimension even though the objective possesses an exponential number of saddle points. This efficient convergence can be viewed as a consequence of negative curvature normal to the stable manifolds associated with saddle points, and we provide evidence that this feature is shared by other nonconvex problems of importance as well." @default.
- W2892755733 created "2018-10-05" @default.
- W2892755733 creator A5017296839 @default.
- W2892755733 creator A5019610791 @default.
- W2892755733 creator A5026684245 @default.
- W2892755733 date "2018-09-27" @default.
- W2892755733 modified "2023-09-27" @default.
- W2892755733 title "Efficient Dictionary Learning with Gradient Descent" @default.
- W2892755733 cites W1811750039 @default.
- W2892755733 cites W1850240193 @default.
- W2892755733 cites W1955967056 @default.
- W2892755733 cites W2592651140 @default.
- W2892755733 hasPublicationYear "2018" @default.
- W2892755733 type Work @default.
- W2892755733 sameAs 2892755733 @default.
- W2892755733 citedByCount "4" @default.
- W2892755733 countsByYear W28927557332017 @default.
- W2892755733 countsByYear W28927557332019 @default.
- W2892755733 crossrefType "posted-content" @default.
- W2892755733 hasAuthorship W2892755733A5017296839 @default.
- W2892755733 hasAuthorship W2892755733A5019610791 @default.
- W2892755733 hasAuthorship W2892755733A5026684245 @default.
- W2892755733 hasConcept C11413529 @default.
- W2892755733 hasConcept C114614502 @default.
- W2892755733 hasConcept C116149140 @default.
- W2892755733 hasConcept C126255220 @default.
- W2892755733 hasConcept C127413603 @default.
- W2892755733 hasConcept C134306372 @default.
- W2892755733 hasConcept C146978453 @default.
- W2892755733 hasConcept C151376022 @default.
- W2892755733 hasConcept C153258448 @default.
- W2892755733 hasConcept C154945302 @default.
- W2892755733 hasConcept C162324750 @default.
- W2892755733 hasConcept C195065555 @default.
- W2892755733 hasConcept C2524010 @default.
- W2892755733 hasConcept C2681867 @default.
- W2892755733 hasConcept C2776637919 @default.
- W2892755733 hasConcept C2777127463 @default.
- W2892755733 hasConcept C2777303404 @default.
- W2892755733 hasConcept C28826006 @default.
- W2892755733 hasConcept C33676613 @default.
- W2892755733 hasConcept C33923547 @default.
- W2892755733 hasConcept C41008148 @default.
- W2892755733 hasConcept C50522688 @default.
- W2892755733 hasConcept C50644808 @default.
- W2892755733 hasConceptScore W2892755733C11413529 @default.
- W2892755733 hasConceptScore W2892755733C114614502 @default.
- W2892755733 hasConceptScore W2892755733C116149140 @default.
- W2892755733 hasConceptScore W2892755733C126255220 @default.
- W2892755733 hasConceptScore W2892755733C127413603 @default.
- W2892755733 hasConceptScore W2892755733C134306372 @default.
- W2892755733 hasConceptScore W2892755733C146978453 @default.
- W2892755733 hasConceptScore W2892755733C151376022 @default.
- W2892755733 hasConceptScore W2892755733C153258448 @default.
- W2892755733 hasConceptScore W2892755733C154945302 @default.
- W2892755733 hasConceptScore W2892755733C162324750 @default.
- W2892755733 hasConceptScore W2892755733C195065555 @default.
- W2892755733 hasConceptScore W2892755733C2524010 @default.
- W2892755733 hasConceptScore W2892755733C2681867 @default.
- W2892755733 hasConceptScore W2892755733C2776637919 @default.
- W2892755733 hasConceptScore W2892755733C2777127463 @default.
- W2892755733 hasConceptScore W2892755733C2777303404 @default.
- W2892755733 hasConceptScore W2892755733C28826006 @default.
- W2892755733 hasConceptScore W2892755733C33676613 @default.
- W2892755733 hasConceptScore W2892755733C33923547 @default.
- W2892755733 hasConceptScore W2892755733C41008148 @default.
- W2892755733 hasConceptScore W2892755733C50522688 @default.
- W2892755733 hasConceptScore W2892755733C50644808 @default.
- W2892755733 hasLocation W28927557331 @default.
- W2892755733 hasOpenAccess W2892755733 @default.
- W2892755733 hasPrimaryLocation W28927557331 @default.
- W2892755733 hasRelatedWork W2077695581 @default.
- W2892755733 hasRelatedWork W2156374634 @default.
- W2892755733 hasRelatedWork W2276652794 @default.
- W2892755733 hasRelatedWork W2461188804 @default.
- W2892755733 hasRelatedWork W2624903936 @default.
- W2892755733 hasRelatedWork W2688160009 @default.
- W2892755733 hasRelatedWork W2761921260 @default.
- W2892755733 hasRelatedWork W2778898806 @default.
- W2892755733 hasRelatedWork W2888643664 @default.
- W2892755733 hasRelatedWork W2945624305 @default.
- W2892755733 hasRelatedWork W2963357609 @default.
- W2892755733 hasRelatedWork W2964026281 @default.
- W2892755733 hasRelatedWork W2965291189 @default.
- W2892755733 hasRelatedWork W2969951231 @default.
- W2892755733 hasRelatedWork W2970821902 @default.
- W2892755733 hasRelatedWork W2992415385 @default.
- W2892755733 hasRelatedWork W3030916542 @default.
- W2892755733 hasRelatedWork W3034758333 @default.
- W2892755733 hasRelatedWork W3091782125 @default.
- W2892755733 hasRelatedWork W3166476363 @default.
- W2892755733 isParatext "false" @default.
- W2892755733 isRetracted "false" @default.
- W2892755733 magId "2892755733" @default.
- W2892755733 workType "article" @default.