FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W86146351> ?p ?o ?g. }

• W86146351 endingPage "219" @default.
• W86146351 startingPage "219" @default.
• W86146351 abstract "Many applications require predicting not a just a single variable, but multiple variables that depend on each other. Recent attention has therefore focused on structured prediction methods, which combine the modeling flexibility of graphical models with the ability to employ complex, dependent features typical of traditional classification methods. Especially popular have been conditional random fields (CRFs), which are graphical models of the conditional distribution over outputs given a set of observed features. Unfortunately, parameter estimation in CRFs requires repeated inference, which can be computationally expensive. Complex graphical structures are increasingly desired in practical applications, but then training time often becomes prohibitive. In this thesis, I investigate efficient training methods for conditional random fields with complex graphical structure, focusing on local methods which avoid propagating information globally along the graph. First, I investigate piecewise training, which trains each of a model's factors separately. I present three views of piecewise training: as maximizing the likelihood in a so-called node-split graph, as maximizing the Bethe likelihood with uniform messages, and as generalizing the pseudo-moment matching estimator of Wainwright et al. [2003]. Second, I propose piecewise pseudolikelihood, a hybrid procedure which pseudolikelihood-izes the piecewise likelihood, and is therefore more efficient if the variables have large cardinality. Piecewise pseudolikelihood performs well even on applications in which standard pseudolikelihood performs poorly. Finally, motivated by the connection between piecewise training and BP, I explore training methods using beliefs arising from stopping BP before convergence. I propose a new schedule for message propagation that improves upon the dynamic schedule proposed recently by Elidan et al. [2006], and present suggestive results applying dynamic schedules to the system of equations that combine inference and learning. I also present two novel families of loopy CRFs, which appear as test cases throughout. First is the dynamic CRF, which combines the factorized state representation of dynamic Bayesian networks with the modeling flexibility of conditional models. The second of these is the skip-chain CRF, which models the fact that identical words are likely to have the same label, even if they occur far apart." @default.
• W86146351 created "2016-06-24" @default.
• W86146351 creator A5008354502 @default.
• W86146351 creator A5028501178 @default.
• W86146351 date "2008-01-01" @default.
• W86146351 modified "2023-09-26" @default.
• W86146351 title "Efficient training methods for conditional random fields" @default.
• W86146351 cites W122473599 @default.
• W86146351 cites W147273232 @default.
• W86146351 cites W1499578805 @default.
• W86146351 cites W1513861746 @default.
• W86146351 cites W1516111018 @default.
• W86146351 cites W1525888637 @default.
• W86146351 cites W1528056001 @default.
• W86146351 cites W1528797350 @default.
• W86146351 cites W1530235965 @default.
• W86146351 cites W1537309404 @default.
• W86146351 cites W1545898854 @default.
• W86146351 cites W1601974683 @default.
• W86146351 cites W1604803556 @default.
• W86146351 cites W1623072288 @default.
• W86146351 cites W1632114991 @default.
• W86146351 cites W1636244751 @default.
• W86146351 cites W1640763397 @default.
• W86146351 cites W1651266332 @default.
• W86146351 cites W1714704734 @default.
• W86146351 cites W1732623802 @default.
• W86146351 cites W173906397 @default.
• W86146351 cites W1766290689 @default.
• W86146351 cites W1773803948 @default.
• W86146351 cites W1781547478 @default.
• W86146351 cites W1802388531 @default.
• W86146351 cites W1842236176 @default.
• W86146351 cites W1880262756 @default.
• W86146351 cites W1934019294 @default.
• W86146351 cites W1934021597 @default.
• W86146351 cites W1972950354 @default.
• W86146351 cites W1977970897 @default.
• W86146351 cites W1983599491 @default.
• W86146351 cites W1988995507 @default.
• W86146351 cites W1996430422 @default.
• W86146351 cites W1999595522 @default.
• W86146351 cites W2004915807 @default.
• W86146351 cites W2008652694 @default.
• W86146351 cites W2020294948 @default.
• W86146351 cites W2020999234 @default.
• W86146351 cites W2026129786 @default.
• W86146351 cites W2036516910 @default.
• W86146351 cites W2041522124 @default.
• W86146351 cites W2046932483 @default.
• W86146351 cites W2058839679 @default.
• W86146351 cites W2073522073 @default.
• W86146351 cites W2083195487 @default.
• W86146351 cites W2083875149 @default.
• W86146351 cites W2086240273 @default.
• W86146351 cites W2096044236 @default.
• W86146351 cites W2096765155 @default.
• W86146351 cites W2098678088 @default.
• W86146351 cites W2098921539 @default.
• W86146351 cites W2102667697 @default.
• W86146351 cites W2104029044 @default.
• W86146351 cites W2105644991 @default.
• W86146351 cites W2107188136 @default.
• W86146351 cites W2110575115 @default.
• W86146351 cites W2114220616 @default.
• W86146351 cites W2114521167 @default.
• W86146351 cites W2116064496 @default.
• W86146351 cites W2116410915 @default.
• W86146351 cites W2119224513 @default.
• W86146351 cites W2120340025 @default.
• W86146351 cites W2122853437 @default.
• W86146351 cites W2125838338 @default.
• W86146351 cites W2129031807 @default.
• W86146351 cites W2129142203 @default.
• W86146351 cites W2129340397 @default.
• W86146351 cites W2129712609 @default.
• W86146351 cites W2133236963 @default.
• W86146351 cites W2135094946 @default.
• W86146351 cites W2137650255 @default.
• W86146351 cites W2137813581 @default.
• W86146351 cites W2139686264 @default.
• W86146351 cites W2141099517 @default.
• W86146351 cites W2141732516 @default.
• W86146351 cites W2142276227 @default.
• W86146351 cites W2142500366 @default.
• W86146351 cites W2143349571 @default.
• W86146351 cites W2143458716 @default.
• W86146351 cites W2146143523 @default.
• W86146351 cites W2147071755 @default.
• W86146351 cites W2147102238 @default.
• W86146351 cites W2147880316 @default.
• W86146351 cites W2148160361 @default.
• W86146351 cites W2148877605 @default.
• W86146351 cites W2149660837 @default.
• W86146351 cites W2149760002 @default.
• W86146351 cites W2152455533 @default.
• W86146351 cites W2153939756 @default.
• W86146351 cites W2155925463 @default.

• next