Matches in SemOpenAlex for { <https://semopenalex.org/work/W2227388049> ?p ?o ?g. }
Showing items 1 to 69 of
69
with 100 items per page.
- W2227388049 abstract "With advances in computer and information technology that generate data with increasing volume and complexity, learning from data has become a key challenge. The evolving technology also impacts the way communication networks are designed by enabling all-wireless communication between millimeter-sized sensors. Motivated by these challenges, this thesis considers problems in the areas of statistical learning and wireless communication, with a focus on fundamental limits that provide guideline for practical applications. The first part of this work concentrates on learning algorithms based on greedy minimization of loss functions (e.g., boosting). Our first contribution is the introduction of recursive greedy algorithms. While the standard batch method as the sample size increases is to re-initialize the greedy algorithm with an arbitrary rule, the proposed procedures proceed with a composite classifier obtained earlier for smaller sample size. We prove that the recursive methods are Bayes-consistent. Experiments demonstrate the practical benefits of these methods for the practitioner who continually receives new batches of training examples or has to cope with large data sets. Another principal finding is to generalize consistency results for boosting methods originally obtained under the assumption of independent observations to the much less restrictive case of weakly dependent observations. Our investigation is motivated by the fact that in practice observations are rarely independent; ignoring dependence can seriously undermine performance. We obtain a consistency result in which the less restricted nature of sampling is manifested through a generalized condition on the growth of a regularization parameter. The last part of the thesis concerns fundamental limits of all-wireless networks. We concentrate on the scaling of the throughput capacity. Previous results indicate that with an increasing number of nodes, the throughput collapses to zero for immobile nodes, while it can be kept constant if the nodes move freely in the communication domain. We analyze the impact of restricting mobility on throughput scaling, and obtain a general throughput result which is a function of simple properties of the network. It is shown to capture every order of growth for the throughput, encompassing the results for immobile and fully mobile nodes as extremes." @default.
- W2227388049 created "2016-06-24" @default.
- W2227388049 creator A5055054658 @default.
- W2227388049 date "2007-01-01" @default.
- W2227388049 modified "2023-09-23" @default.
- W2227388049 title "Asymptotics of boosting, greedy learning algorithms, and wireless networks" @default.
- W2227388049 hasPublicationYear "2007" @default.
- W2227388049 type Work @default.
- W2227388049 sameAs 2227388049 @default.
- W2227388049 citedByCount "0" @default.
- W2227388049 crossrefType "journal-article" @default.
- W2227388049 hasAuthorship W2227388049A5055054658 @default.
- W2227388049 hasConcept C105795698 @default.
- W2227388049 hasConcept C11413529 @default.
- W2227388049 hasConcept C119857082 @default.
- W2227388049 hasConcept C129848803 @default.
- W2227388049 hasConcept C147764199 @default.
- W2227388049 hasConcept C154945302 @default.
- W2227388049 hasConcept C199360897 @default.
- W2227388049 hasConcept C2776436953 @default.
- W2227388049 hasConcept C33923547 @default.
- W2227388049 hasConcept C41008148 @default.
- W2227388049 hasConcept C46686674 @default.
- W2227388049 hasConcept C51823790 @default.
- W2227388049 hasConcept C555944384 @default.
- W2227388049 hasConcept C76155785 @default.
- W2227388049 hasConcept C95623464 @default.
- W2227388049 hasConceptScore W2227388049C105795698 @default.
- W2227388049 hasConceptScore W2227388049C11413529 @default.
- W2227388049 hasConceptScore W2227388049C119857082 @default.
- W2227388049 hasConceptScore W2227388049C129848803 @default.
- W2227388049 hasConceptScore W2227388049C147764199 @default.
- W2227388049 hasConceptScore W2227388049C154945302 @default.
- W2227388049 hasConceptScore W2227388049C199360897 @default.
- W2227388049 hasConceptScore W2227388049C2776436953 @default.
- W2227388049 hasConceptScore W2227388049C33923547 @default.
- W2227388049 hasConceptScore W2227388049C41008148 @default.
- W2227388049 hasConceptScore W2227388049C46686674 @default.
- W2227388049 hasConceptScore W2227388049C51823790 @default.
- W2227388049 hasConceptScore W2227388049C555944384 @default.
- W2227388049 hasConceptScore W2227388049C76155785 @default.
- W2227388049 hasConceptScore W2227388049C95623464 @default.
- W2227388049 hasLocation W22273880491 @default.
- W2227388049 hasOpenAccess W2227388049 @default.
- W2227388049 hasPrimaryLocation W22273880491 @default.
- W2227388049 hasRelatedWork W1784081702 @default.
- W2227388049 hasRelatedWork W1971648331 @default.
- W2227388049 hasRelatedWork W2000523271 @default.
- W2227388049 hasRelatedWork W2107292930 @default.
- W2227388049 hasRelatedWork W2135333267 @default.
- W2227388049 hasRelatedWork W2171548678 @default.
- W2227388049 hasRelatedWork W2336405987 @default.
- W2227388049 hasRelatedWork W268427563 @default.
- W2227388049 hasRelatedWork W2775077303 @default.
- W2227388049 hasRelatedWork W2799013664 @default.
- W2227388049 hasRelatedWork W2972428370 @default.
- W2227388049 hasRelatedWork W2982154564 @default.
- W2227388049 hasRelatedWork W3028382483 @default.
- W2227388049 hasRelatedWork W3134556739 @default.
- W2227388049 hasRelatedWork W3158156290 @default.
- W2227388049 hasRelatedWork W36041826 @default.
- W2227388049 hasRelatedWork W2136943590 @default.
- W2227388049 hasRelatedWork W2508181126 @default.
- W2227388049 hasRelatedWork W2553384521 @default.
- W2227388049 hasRelatedWork W3126879498 @default.
- W2227388049 isParatext "false" @default.
- W2227388049 isRetracted "false" @default.
- W2227388049 magId "2227388049" @default.
- W2227388049 workType "article" @default.