FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W1785095627> ?p ?o ?g. }

• W1785095627 endingPage "119" @default.
• W1785095627 startingPage "119" @default.
• W1785095627 abstract "The connection between the conditioning of a problem instance—the sensitivity of a problem instance to perturbations in the input—and the speed of certain iterative algorithms in solving that problem instance is a recurring topic of study in numerical analysis. This dissertation, consisting of three distinct parts, provides a further connection through the framework of randomized optimization algorithms. In Part I, we explore how randomization can help asymptotic convergence properties of simple, directional search-based optimization methods. Specifically, we develop a randomized, iterative scheme for estimating the Hessian matrix of a twice-differentiable function. Using this estimation technique, we analyze how it can be used to enhance a random directional search method. From there, we proceed to develop a conjugate-directional search method that incorporates estimated Hessian information without requiring direct use of gradients. In Part II, we turn our focus to randomized variants of two classical algorithms: coordinate descent methods for systems of linear equations and iterated projection methods for systems of linear inequalities. We then demonstrate that, under appropriate randomization schemes, linear rates of convergence can be bounded (in expectation) in terms of natural linear-algebraic conditioning measures for these problems. By considering conditioning concepts induced by metric regularity and metric subregularity, we then expand upon these results by examining randomized projection algorithms for convex feasibility problems. Extensions to reflection-based algorithms are also discussed. Observing that convex feasibility problems can be reformulated into the problem of finding a common zero of maximal monotone operators, we proceed by studying the proximal point method in Part III. Specifically, for the problem of finding a zero of a single maximal monotone operator, we show that metric subregularity of that operator is sufficient for linear convergence of the proximal point method, leading to a convergence rate in terms of the conditioning induced by the modulus of subregularity. This result is then generalized—by considering randomized and averaged proximal point methods—to obtain a convergence rate for the problem of finding a common zero of finitely many such operators." @default.
• W1785095627 created "2016-06-24" @default.
• W1785095627 creator A5038517865 @default.
• W1785095627 creator A5081400874 @default.
• W1785095627 date "2009-01-01" @default.
• W1785095627 modified "2023-09-24" @default.
• W1785095627 title "Effects of conditioning on the convergence of randomized optimization algorithms" @default.
• W1785095627 cites W135911866 @default.
• W1785095627 cites W1487976166 @default.
• W1785095627 cites W1488435683 @default.
• W1785095627 cites W1495915592 @default.
• W1785095627 cites W1497941563 @default.
• W1785095627 cites W1503425191 @default.
• W1785095627 cites W1509803206 @default.
• W1785095627 cites W1525779601 @default.
• W1785095627 cites W1537526063 @default.
• W1785095627 cites W1632219083 @default.
• W1785095627 cites W1885121406 @default.
• W1785095627 cites W1960239420 @default.
• W1785095627 cites W1964995199 @default.
• W1785095627 cites W1966169391 @default.
• W1785095627 cites W1974531281 @default.
• W1785095627 cites W1975552883 @default.
• W1785095627 cites W1979089199 @default.
• W1785095627 cites W1979750072 @default.
• W1785095627 cites W1983871704 @default.
• W1785095627 cites W1987898022 @default.
• W1785095627 cites W1991203806 @default.
• W1785095627 cites W1991607307 @default.
• W1785095627 cites W1992228381 @default.
• W1785095627 cites W1992386684 @default.
• W1785095627 cites W2002728800 @default.
• W1785095627 cites W2004001705 @default.
• W1785095627 cites W2004026774 @default.
• W1785095627 cites W2005045813 @default.
• W1785095627 cites W2008195537 @default.
• W1785095627 cites W2009843399 @default.
• W1785095627 cites W2010003677 @default.
• W1785095627 cites W2012062489 @default.
• W1785095627 cites W2013548809 @default.
• W1785095627 cites W2013850411 @default.
• W1785095627 cites W2020004242 @default.
• W1785095627 cites W2020708563 @default.
• W1785095627 cites W2020991188 @default.
• W1785095627 cites W2026573098 @default.
• W1785095627 cites W2028273959 @default.
• W1785095627 cites W2028324692 @default.
• W1785095627 cites W203276351 @default.
• W1785095627 cites W2033316062 @default.
• W1785095627 cites W2033511209 @default.
• W1785095627 cites W2033604629 @default.
• W1785095627 cites W2035814233 @default.
• W1785095627 cites W2038497950 @default.
• W1785095627 cites W2038998663 @default.
• W1785095627 cites W2042261731 @default.
• W1785095627 cites W2042465463 @default.
• W1785095627 cites W2043804332 @default.
• W1785095627 cites W2045152979 @default.
• W1785095627 cites W2047341913 @default.
• W1785095627 cites W2048305372 @default.
• W1785095627 cites W2052059821 @default.
• W1785095627 cites W2054543516 @default.
• W1785095627 cites W2056099894 @default.
• W1785095627 cites W2062059813 @default.
• W1785095627 cites W2063685783 @default.
• W1785095627 cites W2064076655 @default.
• W1785095627 cites W2065114871 @default.
• W1785095627 cites W2070073065 @default.
• W1785095627 cites W2072781311 @default.
• W1785095627 cites W2074214991 @default.
• W1785095627 cites W2079004067 @default.
• W1785095627 cites W2082451773 @default.
• W1785095627 cites W2083887146 @default.
• W1785095627 cites W2083894543 @default.
• W1785095627 cites W2085440994 @default.
• W1785095627 cites W2089028133 @default.
• W1785095627 cites W2092335343 @default.
• W1785095627 cites W2093417350 @default.
• W1785095627 cites W2094508696 @default.
• W1785095627 cites W2099679613 @default.
• W1785095627 cites W2100904599 @default.
• W1785095627 cites W2101345238 @default.
• W1785095627 cites W2101769813 @default.
• W1785095627 cites W2109154566 @default.
• W1785095627 cites W2114013702 @default.
• W1785095627 cites W2116174279 @default.
• W1785095627 cites W2116598569 @default.
• W1785095627 cites W2117744671 @default.
• W1785095627 cites W2118152326 @default.
• W1785095627 cites W2123859926 @default.
• W1785095627 cites W2132614631 @default.
• W1785095627 cites W2135294617 @default.
• W1785095627 cites W2148609440 @default.
• W1785095627 cites W2150140216 @default.
• W1785095627 cites W2158902838 @default.
• W1785095627 cites W2160068964 @default.
• W1785095627 cites W2160701402 @default.
• W1785095627 cites W2166406271 @default.

• next