Matches in SemOpenAlex for { <https://semopenalex.org/work/W3049275070> ?p ?o ?g. }
- W3049275070 abstract "We provide new results for noise-tolerant and sample-efficient learning algorithms under $s$-concave distributions. The new class of $s$-concave distributions is a broad and natural generalization of log-concavity, and includes many important additional distributions, e.g., the Pareto distribution and $t$-distribution. This class has been studied in the context of efficient sampling, integration, and optimization, but much remains unknown about the geometry of this class of distributions and their applications in the context of learning. The challenge is that unlike the commonly used distributions in learning (uniform or more generally log-concave distributions), this broader class is not closed under the marginalization operator and many such distributions are fat-tailed. In this work, we introduce new convex geometry tools to study the properties of $s$-concave distributions and use these properties to provide bounds on quantities of interest to learning including the probability of disagreement between two halfspaces, disagreement outside a band, and the disagreement coefficient. We use these results to significantly generalize prior results for margin-based active learning, disagreement-based active learning, and passive learning of intersections of halfspaces. Our analysis of geometric properties of $s$-concave distributions might be of independent interest to optimization more broadly." @default.
- W3049275070 created "2020-08-21" @default.
- W3049275070 creator A5018348928 @default.
- W3049275070 creator A5068544954 @default.
- W3049275070 date "2017-03-22" @default.
- W3049275070 modified "2023-09-27" @default.
- W3049275070 title "Sample and Computationally Efficient Learning Algorithms under S-Concave Distributions" @default.
- W3049275070 cites W115950659 @default.
- W3049275070 cites W1502281788 @default.
- W3049275070 cites W1520252399 @default.
- W3049275070 cites W1525870773 @default.
- W3049275070 cites W1542886316 @default.
- W3049275070 cites W1955298890 @default.
- W3049275070 cites W1965187358 @default.
- W3049275070 cites W1968998685 @default.
- W3049275070 cites W1999830310 @default.
- W3049275070 cites W2056138823 @default.
- W3049275070 cites W2056707879 @default.
- W3049275070 cites W2065923584 @default.
- W3049275070 cites W2069691149 @default.
- W3049275070 cites W2097373287 @default.
- W3049275070 cites W2101930583 @default.
- W3049275070 cites W2106318612 @default.
- W3049275070 cites W2106458073 @default.
- W3049275070 cites W2109336881 @default.
- W3049275070 cites W2110327402 @default.
- W3049275070 cites W2110503949 @default.
- W3049275070 cites W2114232233 @default.
- W3049275070 cites W2117756453 @default.
- W3049275070 cites W2121541286 @default.
- W3049275070 cites W2128518360 @default.
- W3049275070 cites W2129113961 @default.
- W3049275070 cites W2140125566 @default.
- W3049275070 cites W2143362693 @default.
- W3049275070 cites W2143521632 @default.
- W3049275070 cites W2151023586 @default.
- W3049275070 cites W2151596305 @default.
- W3049275070 cites W2154952480 @default.
- W3049275070 cites W2156909104 @default.
- W3049275070 cites W2167595980 @default.
- W3049275070 cites W2168908795 @default.
- W3049275070 cites W2276364677 @default.
- W3049275070 cites W2475016247 @default.
- W3049275070 cites W2591805537 @default.
- W3049275070 cites W2611121736 @default.
- W3049275070 cites W28787979 @default.
- W3049275070 cites W2951528191 @default.
- W3049275070 cites W2951563626 @default.
- W3049275070 cites W2962703558 @default.
- W3049275070 cites W2963402790 @default.
- W3049275070 cites W2963537613 @default.
- W3049275070 cites W3021023808 @default.
- W3049275070 cites W43111499 @default.
- W3049275070 hasPublicationYear "2017" @default.
- W3049275070 type Work @default.
- W3049275070 sameAs 3049275070 @default.
- W3049275070 citedByCount "2" @default.
- W3049275070 countsByYear W30492750702017 @default.
- W3049275070 crossrefType "posted-content" @default.
- W3049275070 hasAuthorship W3049275070A5018348928 @default.
- W3049275070 hasAuthorship W3049275070A5068544954 @default.
- W3049275070 hasConcept C105795698 @default.
- W3049275070 hasConcept C110121322 @default.
- W3049275070 hasConcept C112680207 @default.
- W3049275070 hasConcept C11413529 @default.
- W3049275070 hasConcept C119857082 @default.
- W3049275070 hasConcept C126255220 @default.
- W3049275070 hasConcept C128963836 @default.
- W3049275070 hasConcept C134306372 @default.
- W3049275070 hasConcept C149441793 @default.
- W3049275070 hasConcept C151730666 @default.
- W3049275070 hasConcept C154945302 @default.
- W3049275070 hasConcept C177148314 @default.
- W3049275070 hasConcept C185592680 @default.
- W3049275070 hasConcept C198531522 @default.
- W3049275070 hasConcept C2524010 @default.
- W3049275070 hasConcept C2777212361 @default.
- W3049275070 hasConcept C2779343474 @default.
- W3049275070 hasConcept C28826006 @default.
- W3049275070 hasConcept C33923547 @default.
- W3049275070 hasConcept C41008148 @default.
- W3049275070 hasConcept C43617362 @default.
- W3049275070 hasConcept C774472 @default.
- W3049275070 hasConcept C86803240 @default.
- W3049275070 hasConceptScore W3049275070C105795698 @default.
- W3049275070 hasConceptScore W3049275070C110121322 @default.
- W3049275070 hasConceptScore W3049275070C112680207 @default.
- W3049275070 hasConceptScore W3049275070C11413529 @default.
- W3049275070 hasConceptScore W3049275070C119857082 @default.
- W3049275070 hasConceptScore W3049275070C126255220 @default.
- W3049275070 hasConceptScore W3049275070C128963836 @default.
- W3049275070 hasConceptScore W3049275070C134306372 @default.
- W3049275070 hasConceptScore W3049275070C149441793 @default.
- W3049275070 hasConceptScore W3049275070C151730666 @default.
- W3049275070 hasConceptScore W3049275070C154945302 @default.
- W3049275070 hasConceptScore W3049275070C177148314 @default.
- W3049275070 hasConceptScore W3049275070C185592680 @default.
- W3049275070 hasConceptScore W3049275070C198531522 @default.
- W3049275070 hasConceptScore W3049275070C2524010 @default.
- W3049275070 hasConceptScore W3049275070C2777212361 @default.