Matches in SemOpenAlex for { <https://semopenalex.org/work/W3012571504> ?p ?o ?g. }
- W3012571504 abstract "Modern variational inference (VI) uses stochastic gradients to avoid intractable expectations, enabling large-scale probabilistic inference in complex models. VI posits a family of approximating distributions q and then finds the member of that family that is closest to the exact posterior p. Traditionally, VI algorithms minimize the exclusive Kullback-Leibler (KL) KL(q || p), often for computational convenience. Recent research, however, has also focused on the KL(p || q), which has good statistical properties that makes it more appropriate for certain inference problems. This paper develops a simple algorithm for reliably minimizing the inclusive KL using stochastic gradients with vanishing bias. This method, which we call Markovian score climbing (MSC), converges to a local optimum of the inclusive KL. It does not suffer from the systematic errors inherent in existing methods, such as Reweighted Wake-Sleep and Neural Adaptive Sequential Monte Carlo, which lead to bias in their final estimates. We illustrate convergence on a toy model and demonstrate the utility of MSC on Bayesian probit regression for classification as well as a stochastic volatility model for financial data." @default.
- W3012571504 created "2020-03-27" @default.
- W3012571504 creator A5020493508 @default.
- W3012571504 creator A5045048407 @default.
- W3012571504 creator A5070920982 @default.
- W3012571504 date "2020-03-23" @default.
- W3012571504 modified "2023-09-27" @default.
- W3012571504 title "Markovian Score Climbing: Variational Inference with KL(p||q)" @default.
- W3012571504 cites W1499021337 @default.
- W3012571504 cites W1501586228 @default.
- W3012571504 cites W1516111018 @default.
- W3012571504 cites W1522301498 @default.
- W3012571504 cites W1530235965 @default.
- W3012571504 cites W1534417711 @default.
- W3012571504 cites W1568229137 @default.
- W3012571504 cites W1812949503 @default.
- W3012571504 cites W1824523713 @default.
- W3012571504 cites W1906598733 @default.
- W3012571504 cites W1934021597 @default.
- W3012571504 cites W1959608418 @default.
- W3012571504 cites W1979969656 @default.
- W3012571504 cites W1985093013 @default.
- W3012571504 cites W1994616650 @default.
- W3012571504 cites W2043073960 @default.
- W3012571504 cites W2055742299 @default.
- W3012571504 cites W2098084154 @default.
- W3012571504 cites W2119196781 @default.
- W3012571504 cites W2140051159 @default.
- W3012571504 cites W2198904917 @default.
- W3012571504 cites W2397262406 @default.
- W3012571504 cites W2594807464 @default.
- W3012571504 cites W2735102987 @default.
- W3012571504 cites W2739517340 @default.
- W3012571504 cites W2809829874 @default.
- W3012571504 cites W2885659818 @default.
- W3012571504 cites W2890010095 @default.
- W3012571504 cites W2902483955 @default.
- W3012571504 cites W2906762693 @default.
- W3012571504 cites W2920868047 @default.
- W3012571504 cites W2944347733 @default.
- W3012571504 cites W2951493172 @default.
- W3012571504 cites W2951999424 @default.
- W3012571504 cites W2962897886 @default.
- W3012571504 cites W2963169412 @default.
- W3012571504 cites W2963173382 @default.
- W3012571504 cites W2963275229 @default.
- W3012571504 cites W2963319343 @default.
- W3012571504 cites W2963383053 @default.
- W3012571504 cites W2963723789 @default.
- W3012571504 cites W2963828071 @default.
- W3012571504 cites W2963932270 @default.
- W3012571504 cites W2963936947 @default.
- W3012571504 cites W2963942415 @default.
- W3012571504 cites W2964260519 @default.
- W3012571504 cites W2964318004 @default.
- W3012571504 cites W2964334616 @default.
- W3012571504 cites W2969896603 @default.
- W3012571504 cites W2970799057 @default.
- W3012571504 cites W2973674981 @default.
- W3012571504 cites W2974452757 @default.
- W3012571504 cites W2990572911 @default.
- W3012571504 cites W3091044728 @default.
- W3012571504 cites W3104819538 @default.
- W3012571504 cites W3106193260 @default.
- W3012571504 cites W3120740533 @default.
- W3012571504 hasPublicationYear "2020" @default.
- W3012571504 type Work @default.
- W3012571504 sameAs 3012571504 @default.
- W3012571504 citedByCount "6" @default.
- W3012571504 countsByYear W30125715042019 @default.
- W3012571504 countsByYear W30125715042020 @default.
- W3012571504 countsByYear W30125715042021 @default.
- W3012571504 crossrefType "posted-content" @default.
- W3012571504 hasAuthorship W3012571504A5020493508 @default.
- W3012571504 hasAuthorship W3012571504A5045048407 @default.
- W3012571504 hasAuthorship W3012571504A5070920982 @default.
- W3012571504 hasConcept C107673813 @default.
- W3012571504 hasConcept C111350023 @default.
- W3012571504 hasConcept C11413529 @default.
- W3012571504 hasConcept C126255220 @default.
- W3012571504 hasConcept C149782125 @default.
- W3012571504 hasConcept C154945302 @default.
- W3012571504 hasConcept C160234255 @default.
- W3012571504 hasConcept C162376815 @default.
- W3012571504 hasConcept C2776214188 @default.
- W3012571504 hasConcept C33923547 @default.
- W3012571504 hasConcept C41008148 @default.
- W3012571504 hasConceptScore W3012571504C107673813 @default.
- W3012571504 hasConceptScore W3012571504C111350023 @default.
- W3012571504 hasConceptScore W3012571504C11413529 @default.
- W3012571504 hasConceptScore W3012571504C126255220 @default.
- W3012571504 hasConceptScore W3012571504C149782125 @default.
- W3012571504 hasConceptScore W3012571504C154945302 @default.
- W3012571504 hasConceptScore W3012571504C160234255 @default.
- W3012571504 hasConceptScore W3012571504C162376815 @default.
- W3012571504 hasConceptScore W3012571504C2776214188 @default.
- W3012571504 hasConceptScore W3012571504C33923547 @default.
- W3012571504 hasConceptScore W3012571504C41008148 @default.
- W3012571504 hasLocation W30125715041 @default.
- W3012571504 hasOpenAccess W3012571504 @default.