Matches in SemOpenAlex for { <https://semopenalex.org/work/W4226503158> ?p ?o ?g. }
Showing items 1 to 77 of
77
with 100 items per page.
- W4226503158 abstract "Overparametrized neural networks tend to perfectly fit noisy training data yet generalize well on test data. Inspired by this empirical observation, recent work has sought to understand this phenomenon of benign overfitting or harmless interpolation in the much simpler linear model. Previous theoretical work critically assumes that either the data features are statistically independent or the input data is high-dimensional; this precludes general nonparametric settings with structured feature maps. In this paper, we present a general and flexible framework for upper bounding regression and classification risk in a reproducing kernel Hilbert space. A key contribution is that our framework describes precise sufficient conditions on the data Gram matrix under which harmless interpolation occurs. Our results recover prior independent-features results (with a much simpler analysis), but they furthermore show that harmless interpolation can occur in more general settings such as features that are a bounded orthonormal system. Furthermore, our results show an asymptotic separation between classification and regression performance in a manner that was previously only shown for Gaussian features." @default.
- W4226503158 created "2022-05-05" @default.
- W4226503158 creator A5061744441 @default.
- W4226503158 creator A5064232831 @default.
- W4226503158 creator A5069307903 @default.
- W4226503158 creator A5087204171 @default.
- W4226503158 date "2021-11-09" @default.
- W4226503158 modified "2023-09-29" @default.
- W4226503158 title "Harmless interpolation in regression and classification with structured features" @default.
- W4226503158 doi "https://doi.org/10.48550/arxiv.2111.05198" @default.
- W4226503158 hasPublicationYear "2021" @default.
- W4226503158 type Work @default.
- W4226503158 citedByCount "0" @default.
- W4226503158 crossrefType "posted-content" @default.
- W4226503158 hasAuthorship W4226503158A5061744441 @default.
- W4226503158 hasAuthorship W4226503158A5064232831 @default.
- W4226503158 hasAuthorship W4226503158A5069307903 @default.
- W4226503158 hasAuthorship W4226503158A5087204171 @default.
- W4226503158 hasBestOaLocation W42265031581 @default.
- W4226503158 hasConcept C102366305 @default.
- W4226503158 hasConcept C104114177 @default.
- W4226503158 hasConcept C105795698 @default.
- W4226503158 hasConcept C11413529 @default.
- W4226503158 hasConcept C119857082 @default.
- W4226503158 hasConcept C134306372 @default.
- W4226503158 hasConcept C137800194 @default.
- W4226503158 hasConcept C138885662 @default.
- W4226503158 hasConcept C153180895 @default.
- W4226503158 hasConcept C154945302 @default.
- W4226503158 hasConcept C22019652 @default.
- W4226503158 hasConcept C2776401178 @default.
- W4226503158 hasConcept C33923547 @default.
- W4226503158 hasConcept C34388435 @default.
- W4226503158 hasConcept C41008148 @default.
- W4226503158 hasConcept C41895202 @default.
- W4226503158 hasConcept C50644808 @default.
- W4226503158 hasConcept C62799726 @default.
- W4226503158 hasConcept C63584917 @default.
- W4226503158 hasConcept C74127309 @default.
- W4226503158 hasConcept C80884492 @default.
- W4226503158 hasConceptScore W4226503158C102366305 @default.
- W4226503158 hasConceptScore W4226503158C104114177 @default.
- W4226503158 hasConceptScore W4226503158C105795698 @default.
- W4226503158 hasConceptScore W4226503158C11413529 @default.
- W4226503158 hasConceptScore W4226503158C119857082 @default.
- W4226503158 hasConceptScore W4226503158C134306372 @default.
- W4226503158 hasConceptScore W4226503158C137800194 @default.
- W4226503158 hasConceptScore W4226503158C138885662 @default.
- W4226503158 hasConceptScore W4226503158C153180895 @default.
- W4226503158 hasConceptScore W4226503158C154945302 @default.
- W4226503158 hasConceptScore W4226503158C22019652 @default.
- W4226503158 hasConceptScore W4226503158C2776401178 @default.
- W4226503158 hasConceptScore W4226503158C33923547 @default.
- W4226503158 hasConceptScore W4226503158C34388435 @default.
- W4226503158 hasConceptScore W4226503158C41008148 @default.
- W4226503158 hasConceptScore W4226503158C41895202 @default.
- W4226503158 hasConceptScore W4226503158C50644808 @default.
- W4226503158 hasConceptScore W4226503158C62799726 @default.
- W4226503158 hasConceptScore W4226503158C63584917 @default.
- W4226503158 hasConceptScore W4226503158C74127309 @default.
- W4226503158 hasConceptScore W4226503158C80884492 @default.
- W4226503158 hasLocation W42265031581 @default.
- W4226503158 hasOpenAccess W4226503158 @default.
- W4226503158 hasPrimaryLocation W42265031581 @default.
- W4226503158 hasRelatedWork W2767651786 @default.
- W4226503158 hasRelatedWork W2891726612 @default.
- W4226503158 hasRelatedWork W2985459377 @default.
- W4226503158 hasRelatedWork W2989932438 @default.
- W4226503158 hasRelatedWork W3099765033 @default.
- W4226503158 hasRelatedWork W3138048262 @default.
- W4226503158 hasRelatedWork W3159087789 @default.
- W4226503158 hasRelatedWork W3175189414 @default.
- W4226503158 hasRelatedWork W4210794429 @default.
- W4226503158 hasRelatedWork W4214844979 @default.
- W4226503158 isParatext "false" @default.
- W4226503158 isRetracted "false" @default.
- W4226503158 workType "article" @default.