FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2288187973> ?p ?o ?g. }

• W2288187973 abstract "In this thesis, we investigate diffusion as an algorithmic and analytic tool in statistics and computer science. We address a question arising from computational linguistics, where we wish to understand the behavior of a network of agents modeled as nodes of a graph that adaptively modify their lexicon using data from their neighbors. By introducing a model of memory and a family of coalescing random walks, we prove that they eventually reach a consensus with probability 1. We study distributed averaging on graphs and devise a distributed algorithm that is based on a diffusion process having two time scales. Addressing the question of routing in a network, we use steady-state diffusions corresponding to electrical flow in a network of resistors for oblivious routing and prove that this scheme performs well under a variety of performance measures. Based on a microscopic view of diffusion as an ensemble of particles executing independent Brownian motions, we develop the fastest currently known algorithm for computing the area of the boundary of a convex set. A similar technique is used to produce samplers for the boundaries of convex sets and smooth hypersurfaces that are the boundaries of open sets in Rn, assuming access to samplers for the interior. These algorithms are motivated by Goodness-of-Fit tests in statistics. The halfplane capacity, a quantity often used to parameterize stochastic processes arising in statistical physics, known as Schramm-Loewner evolutions, is shown to be comparable to a more geometric notion. We analyze a class of natural random walks on a Riemannian manifold, and give bounds on the mixing times in terms of the Cheeger constant and a notion of smoothness that relates the random walk to the metric underlying the manifold. A Markov chain having a stationary distribution that is uniform on the interior of a polytope is developed. This is the first chain whose mixing time is strongly polynomial when initiated in the vicinity of the center of mass. This Markov chain can be interpreted as a random walk on a certain Riemannian manifold. The resulting algorithm for sampling polytopes outperforms known algorithms when the number of constraints is of the same order of magnitude as the dimension. We use a variant of this Markov chain to design a randomized version of Dikin's affine scaling algorithm for linear programming. We provide polynomial-time guarantees which do not exist for Dikin's algorithm. Addressing a question from machine learning, under certain smoothness conditions, we prove that a form of weighted surface area is the limit of the weight of graph cuts in a family of random graphs arising in the context of clustering. This is done by relating both to the amount of diffusion across the surface in question. Addressing a related issue on manifolds, we obtain an upper bound on the annealed entropy of the collection of open subsets of a manifold whose boundaries are well-conditioned. This result leads to an upper bound on the number of random samples needed before it is possible to accurately classify data lying on a manifold." @default.
• W2288187973 created "2016-06-24" @default.
• W2288187973 creator A5005785891 @default.
• W2288187973 creator A5029536777 @default.
• W2288187973 date "2009-01-01" @default.
• W2288187973 modified "2023-09-27" @default.
• W2288187973 title "Diffusion in computer science and statistics" @default.
• W2288187973 hasPublicationYear "2009" @default.
• W2288187973 type Work @default.
• W2288187973 sameAs 2288187973 @default.
• W2288187973 citedByCount "0" @default.
• W2288187973 crossrefType "journal-article" @default.
• W2288187973 hasAuthorship W2288187973A5005785891 @default.
• W2288187973 hasAuthorship W2288187973A5029536777 @default.
• W2288187973 hasConcept C105795698 @default.
• W2288187973 hasConcept C112401455 @default.
• W2288187973 hasConcept C11413529 @default.
• W2288187973 hasConcept C121194460 @default.
• W2288187973 hasConcept C121332964 @default.
• W2288187973 hasConcept C121864883 @default.
• W2288187973 hasConcept C134306372 @default.
• W2288187973 hasConcept C33923547 @default.
• W2288187973 hasConcept C41008148 @default.
• W2288187973 hasConcept C62354387 @default.
• W2288187973 hasConcept C80444323 @default.
• W2288187973 hasConceptScore W2288187973C105795698 @default.
• W2288187973 hasConceptScore W2288187973C112401455 @default.
• W2288187973 hasConceptScore W2288187973C11413529 @default.
• W2288187973 hasConceptScore W2288187973C121194460 @default.
• W2288187973 hasConceptScore W2288187973C121332964 @default.
• W2288187973 hasConceptScore W2288187973C121864883 @default.
• W2288187973 hasConceptScore W2288187973C134306372 @default.
• W2288187973 hasConceptScore W2288187973C33923547 @default.
• W2288187973 hasConceptScore W2288187973C41008148 @default.
• W2288187973 hasConceptScore W2288187973C62354387 @default.
• W2288187973 hasConceptScore W2288187973C80444323 @default.
• W2288187973 hasLocation W22881879731 @default.
• W2288187973 hasOpenAccess W2288187973 @default.
• W2288187973 hasPrimaryLocation W22881879731 @default.
• W2288187973 hasRelatedWork W1703398515 @default.
• W2288187973 hasRelatedWork W1861295442 @default.
• W2288187973 hasRelatedWork W2035598542 @default.
• W2288187973 hasRelatedWork W2043702293 @default.
• W2288187973 hasRelatedWork W2602411525 @default.
• W2288187973 hasRelatedWork W2771044751 @default.
• W2288187973 hasRelatedWork W2798943785 @default.
• W2288187973 hasRelatedWork W2805290903 @default.
• W2288187973 hasRelatedWork W2952534041 @default.
• W2288187973 hasRelatedWork W2970004405 @default.
• W2288187973 hasRelatedWork W3005391585 @default.
• W2288187973 hasRelatedWork W3006365244 @default.
• W2288187973 hasRelatedWork W3006926186 @default.
• W2288187973 hasRelatedWork W3019595353 @default.
• W2288187973 hasRelatedWork W3035211723 @default.
• W2288187973 hasRelatedWork W3080284981 @default.
• W2288187973 hasRelatedWork W3081188487 @default.
• W2288187973 hasRelatedWork W3134412669 @default.
• W2288187973 hasRelatedWork W3181949156 @default.
• W2288187973 hasRelatedWork W2182044025 @default.
• W2288187973 isParatext "false" @default.
• W2288187973 isRetracted "false" @default.
• W2288187973 magId "2288187973" @default.
• W2288187973 workType "article" @default.