Matches in SemOpenAlex for { <https://semopenalex.org/work/W2104952380> ?p ?o ?g. }
- W2104952380 endingPage "608" @default.
- W2104952380 startingPage "601" @default.
- W2104952380 abstract "We investigate a new, convex relaxation of an expectation-maximization (EM) variant that approximates a standard objective while eliminating local minima. First, a cautionary result is presented, showing that any convex relaxation of EM over hidden variables must give trivial results if any dependence on the missing values is retained. Although this appears to be a strong negative outcome, we then demonstrate how the problem can be bypassed by using equivalence relations instead of value assignments over hidden variables. In particular, we develop new algorithms for estimating exponential conditional models that only require equivalence relation information over the variable values. This reformulation leads to an exact expression for EM variants in a wide range of problems. We then develop a semidefinite relaxation that yields global training by eliminating local minima." @default.
- W2104952380 created "2016-06-24" @default.
- W2104952380 creator A5010575626 @default.
- W2104952380 creator A5043824291 @default.
- W2104952380 date "2007-12-03" @default.
- W2104952380 modified "2023-09-24" @default.
- W2104952380 title "Convex Relaxations of Latent Variable Training" @default.
- W2104952380 cites W182881619 @default.
- W2104952380 cites W1976241232 @default.
- W2104952380 cites W1985123706 @default.
- W2104952380 cites W2049633694 @default.
- W2104952380 cites W2102416536 @default.
- W2104952380 cites W2112861996 @default.
- W2104952380 cites W2120340025 @default.
- W2104952380 cites W2132820034 @default.
- W2104952380 cites W2133026689 @default.
- W2104952380 cites W2153568660 @default.
- W2104952380 cites W2155042384 @default.
- W2104952380 cites W2168772685 @default.
- W2104952380 cites W2296319761 @default.
- W2104952380 cites W2567948266 @default.
- W2104952380 cites W3029645440 @default.
- W2104952380 hasPublicationYear "2007" @default.
- W2104952380 type Work @default.
- W2104952380 sameAs 2104952380 @default.
- W2104952380 citedByCount "19" @default.
- W2104952380 countsByYear W21049523802012 @default.
- W2104952380 countsByYear W21049523802013 @default.
- W2104952380 countsByYear W21049523802014 @default.
- W2104952380 countsByYear W21049523802015 @default.
- W2104952380 countsByYear W21049523802016 @default.
- W2104952380 countsByYear W21049523802017 @default.
- W2104952380 countsByYear W21049523802018 @default.
- W2104952380 countsByYear W21049523802019 @default.
- W2104952380 crossrefType "proceedings-article" @default.
- W2104952380 hasAuthorship W2104952380A5010575626 @default.
- W2104952380 hasAuthorship W2104952380A5043824291 @default.
- W2104952380 hasConcept C105795698 @default.
- W2104952380 hasConcept C112680207 @default.
- W2104952380 hasConcept C11413529 @default.
- W2104952380 hasConcept C118615104 @default.
- W2104952380 hasConcept C121332964 @default.
- W2104952380 hasConcept C126255220 @default.
- W2104952380 hasConcept C134306372 @default.
- W2104952380 hasConcept C137002209 @default.
- W2104952380 hasConcept C15744967 @default.
- W2104952380 hasConcept C159985019 @default.
- W2104952380 hasConcept C182365436 @default.
- W2104952380 hasConcept C186633575 @default.
- W2104952380 hasConcept C192562407 @default.
- W2104952380 hasConcept C204323151 @default.
- W2104952380 hasConcept C2524010 @default.
- W2104952380 hasConcept C2776029896 @default.
- W2104952380 hasConcept C2776330181 @default.
- W2104952380 hasConcept C2780069185 @default.
- W2104952380 hasConcept C28826006 @default.
- W2104952380 hasConcept C33923547 @default.
- W2104952380 hasConcept C51167844 @default.
- W2104952380 hasConcept C62520636 @default.
- W2104952380 hasConcept C77805123 @default.
- W2104952380 hasConcept C84114770 @default.
- W2104952380 hasConceptScore W2104952380C105795698 @default.
- W2104952380 hasConceptScore W2104952380C112680207 @default.
- W2104952380 hasConceptScore W2104952380C11413529 @default.
- W2104952380 hasConceptScore W2104952380C118615104 @default.
- W2104952380 hasConceptScore W2104952380C121332964 @default.
- W2104952380 hasConceptScore W2104952380C126255220 @default.
- W2104952380 hasConceptScore W2104952380C134306372 @default.
- W2104952380 hasConceptScore W2104952380C137002209 @default.
- W2104952380 hasConceptScore W2104952380C15744967 @default.
- W2104952380 hasConceptScore W2104952380C159985019 @default.
- W2104952380 hasConceptScore W2104952380C182365436 @default.
- W2104952380 hasConceptScore W2104952380C186633575 @default.
- W2104952380 hasConceptScore W2104952380C192562407 @default.
- W2104952380 hasConceptScore W2104952380C204323151 @default.
- W2104952380 hasConceptScore W2104952380C2524010 @default.
- W2104952380 hasConceptScore W2104952380C2776029896 @default.
- W2104952380 hasConceptScore W2104952380C2776330181 @default.
- W2104952380 hasConceptScore W2104952380C2780069185 @default.
- W2104952380 hasConceptScore W2104952380C28826006 @default.
- W2104952380 hasConceptScore W2104952380C33923547 @default.
- W2104952380 hasConceptScore W2104952380C51167844 @default.
- W2104952380 hasConceptScore W2104952380C62520636 @default.
- W2104952380 hasConceptScore W2104952380C77805123 @default.
- W2104952380 hasConceptScore W2104952380C84114770 @default.
- W2104952380 hasLocation W21049523801 @default.
- W2104952380 hasOpenAccess W2104952380 @default.
- W2104952380 hasPrimaryLocation W21049523801 @default.
- W2104952380 hasRelatedWork W1479807131 @default.
- W2104952380 hasRelatedWork W158976495 @default.
- W2104952380 hasRelatedWork W1615473998 @default.
- W2104952380 hasRelatedWork W1963920598 @default.
- W2104952380 hasRelatedWork W2049633694 @default.
- W2104952380 hasRelatedWork W2086052791 @default.
- W2104952380 hasRelatedWork W2099987944 @default.
- W2104952380 hasRelatedWork W2107008379 @default.
- W2104952380 hasRelatedWork W2107822587 @default.
- W2104952380 hasRelatedWork W2108282816 @default.