Matches in SemOpenAlex for { <https://semopenalex.org/work/W2124905705> ?p ?o ?g. }
Showing items 1 to 90 of
90
with 100 items per page.
- W2124905705 endingPage "2446" @default.
- W2124905705 startingPage "2438" @default.
- W2124905705 abstract "Stochastic variational inference (SVI) lets us scale up Bayesian computation to massive data. It uses stochastic optimization to fit a variational distribution, following easy-to-compute noisy natural gradients. As with most traditional stochastic optimization methods, SVI takes precautions to use unbiased stochastic gradients whose expectations are equal to the true gradients. In this paper, we explore the idea of following biased stochastic gradients in SVI. Our method replaces the natural gradient with a similarly constructed vector that uses a fixed-window moving average of some of its previous terms. We will demonstrate the many advantages of this technique. First, its computational cost is the same as for SVI and storage requirements only multiply by a constant factor. Second, it enjoys significant variance reduction over the unbiased estimates, smaller bias than averaged gradients, and leads to smaller mean-squared error against the full gradient. We test our method on latent Dirichlet allocation with three large corpora." @default.
- W2124905705 created "2016-06-24" @default.
- W2124905705 creator A5036302820 @default.
- W2124905705 creator A5070920982 @default.
- W2124905705 date "2014-12-08" @default.
- W2124905705 modified "2023-10-18" @default.
- W2124905705 title "Smoothed Gradients for Stochastic Variational Inference" @default.
- W2124905705 cites W1506806321 @default.
- W2124905705 cites W1663973292 @default.
- W2124905705 cites W1777124189 @default.
- W2124905705 cites W1880262756 @default.
- W2124905705 cites W1994616650 @default.
- W2124905705 cites W2061863621 @default.
- W2124905705 cites W2066828202 @default.
- W2124905705 cites W2107107106 @default.
- W2124905705 cites W2107438106 @default.
- W2124905705 cites W2120340025 @default.
- W2124905705 cites W2123528936 @default.
- W2124905705 cites W2145832734 @default.
- W2124905705 cites W2165599843 @default.
- W2124905705 cites W2166851633 @default.
- W2124905705 cites W2169713291 @default.
- W2124905705 cites W2171911691 @default.
- W2124905705 cites W2963156201 @default.
- W2124905705 cites W86282824 @default.
- W2124905705 hasPublicationYear "2014" @default.
- W2124905705 type Work @default.
- W2124905705 sameAs 2124905705 @default.
- W2124905705 citedByCount "8" @default.
- W2124905705 countsByYear W21249057052015 @default.
- W2124905705 countsByYear W21249057052016 @default.
- W2124905705 countsByYear W21249057052017 @default.
- W2124905705 countsByYear W21249057052018 @default.
- W2124905705 crossrefType "proceedings-article" @default.
- W2124905705 hasAuthorship W2124905705A5036302820 @default.
- W2124905705 hasAuthorship W2124905705A5070920982 @default.
- W2124905705 hasConcept C11413529 @default.
- W2124905705 hasConcept C126255220 @default.
- W2124905705 hasConcept C134306372 @default.
- W2124905705 hasConcept C154945302 @default.
- W2124905705 hasConcept C169214877 @default.
- W2124905705 hasConcept C182310444 @default.
- W2124905705 hasConcept C194387892 @default.
- W2124905705 hasConcept C2776214188 @default.
- W2124905705 hasConcept C28826006 @default.
- W2124905705 hasConcept C33923547 @default.
- W2124905705 hasConcept C41008148 @default.
- W2124905705 hasConcept C45374587 @default.
- W2124905705 hasConceptScore W2124905705C11413529 @default.
- W2124905705 hasConceptScore W2124905705C126255220 @default.
- W2124905705 hasConceptScore W2124905705C134306372 @default.
- W2124905705 hasConceptScore W2124905705C154945302 @default.
- W2124905705 hasConceptScore W2124905705C169214877 @default.
- W2124905705 hasConceptScore W2124905705C182310444 @default.
- W2124905705 hasConceptScore W2124905705C194387892 @default.
- W2124905705 hasConceptScore W2124905705C2776214188 @default.
- W2124905705 hasConceptScore W2124905705C28826006 @default.
- W2124905705 hasConceptScore W2124905705C33923547 @default.
- W2124905705 hasConceptScore W2124905705C41008148 @default.
- W2124905705 hasConceptScore W2124905705C45374587 @default.
- W2124905705 hasLocation W21249057051 @default.
- W2124905705 hasOpenAccess W2124905705 @default.
- W2124905705 hasPrimaryLocation W21249057051 @default.
- W2124905705 hasRelatedWork W1509040904 @default.
- W2124905705 hasRelatedWork W2047230031 @default.
- W2124905705 hasRelatedWork W2146502635 @default.
- W2124905705 hasRelatedWork W2148721460 @default.
- W2124905705 hasRelatedWork W2166851633 @default.
- W2124905705 hasRelatedWork W2167433878 @default.
- W2124905705 hasRelatedWork W2168947765 @default.
- W2124905705 hasRelatedWork W2418128690 @default.
- W2124905705 hasRelatedWork W2763836425 @default.
- W2124905705 hasRelatedWork W2892457101 @default.
- W2124905705 hasRelatedWork W2909200605 @default.
- W2124905705 hasRelatedWork W2940370615 @default.
- W2124905705 hasRelatedWork W2950413897 @default.
- W2124905705 hasRelatedWork W2963173382 @default.
- W2124905705 hasRelatedWork W2966734222 @default.
- W2124905705 hasRelatedWork W3011226793 @default.
- W2124905705 hasRelatedWork W3027080454 @default.
- W2124905705 hasRelatedWork W3101100385 @default.
- W2124905705 hasRelatedWork W3160612671 @default.
- W2124905705 hasRelatedWork W1569602402 @default.
- W2124905705 hasVolume "27" @default.
- W2124905705 isParatext "false" @default.
- W2124905705 isRetracted "false" @default.
- W2124905705 magId "2124905705" @default.
- W2124905705 workType "article" @default.