Matches in SemOpenAlex for { <https://semopenalex.org/work/W4385474536> ?p ?o ?g. }
Showing items 1 to 89 of
89
with 100 items per page.
- W4385474536 abstract "Deep neural networks (DNNs) trained with the logistic loss (i.e., the cross entropy loss) have made impressive advancements in various binary classification tasks. However, generalization analysis for binary classification with DNNs and logistic loss remains scarce. The unboundedness of the target function for the logistic loss is the main obstacle to deriving satisfying generalization bounds. In this paper, we aim to fill this gap by establishing a novel and elegant oracle-type inequality, which enables us to deal with the boundedness restriction of the target function, and using it to derive sharp convergence rates for fully connected ReLU DNN classifiers trained with logistic loss. In particular, we obtain optimal convergence rates (up to log factors) only requiring the Holder smoothness of the conditional class probability $eta$ of data. Moreover, we consider a compositional assumption that requires $eta$ to be the composition of several vector-valued functions of which each component function is either a maximum value function or a Holder smooth function only depending on a small number of its input variables. Under this assumption, we derive optimal convergence rates (up to log factors) which are independent of the input dimension of data. This result explains why DNN classifiers can perform well in practical high-dimensional classification problems. Besides the novel oracle-type inequality, the sharp convergence rates given in our paper also owe to a tight error bound for approximating the natural logarithm function near zero (where it is unbounded) by ReLU DNNs. In addition, we justify our claims for the optimality of rates by proving corresponding minimax lower bounds. All these results are new in the literature and will deepen our theoretical understanding of classification with DNNs." @default.
- W4385474536 created "2023-08-02" @default.
- W4385474536 creator A5006802589 @default.
- W4385474536 creator A5011681363 @default.
- W4385474536 creator A5027826347 @default.
- W4385474536 date "2023-07-31" @default.
- W4385474536 modified "2023-09-23" @default.
- W4385474536 title "Classification with Deep Neural Networks and Logistic Loss" @default.
- W4385474536 doi "https://doi.org/10.48550/arxiv.2307.16792" @default.
- W4385474536 hasPublicationYear "2023" @default.
- W4385474536 type Work @default.
- W4385474536 citedByCount "0" @default.
- W4385474536 crossrefType "posted-content" @default.
- W4385474536 hasAuthorship W4385474536A5006802589 @default.
- W4385474536 hasAuthorship W4385474536A5011681363 @default.
- W4385474536 hasAuthorship W4385474536A5027826347 @default.
- W4385474536 hasBestOaLocation W43854745361 @default.
- W4385474536 hasConcept C105795698 @default.
- W4385474536 hasConcept C11413529 @default.
- W4385474536 hasConcept C114614502 @default.
- W4385474536 hasConcept C12267149 @default.
- W4385474536 hasConcept C126255220 @default.
- W4385474536 hasConcept C127162648 @default.
- W4385474536 hasConcept C134306372 @default.
- W4385474536 hasConcept C14036430 @default.
- W4385474536 hasConcept C149728462 @default.
- W4385474536 hasConcept C154945302 @default.
- W4385474536 hasConcept C162324750 @default.
- W4385474536 hasConcept C185429906 @default.
- W4385474536 hasConcept C202444582 @default.
- W4385474536 hasConcept C22324862 @default.
- W4385474536 hasConcept C2777303404 @default.
- W4385474536 hasConcept C28826006 @default.
- W4385474536 hasConcept C31258907 @default.
- W4385474536 hasConcept C33676613 @default.
- W4385474536 hasConcept C33923547 @default.
- W4385474536 hasConcept C39927690 @default.
- W4385474536 hasConcept C41008148 @default.
- W4385474536 hasConcept C50522688 @default.
- W4385474536 hasConcept C50644808 @default.
- W4385474536 hasConcept C57869625 @default.
- W4385474536 hasConcept C66905080 @default.
- W4385474536 hasConcept C78458016 @default.
- W4385474536 hasConcept C81388566 @default.
- W4385474536 hasConcept C86803240 @default.
- W4385474536 hasConceptScore W4385474536C105795698 @default.
- W4385474536 hasConceptScore W4385474536C11413529 @default.
- W4385474536 hasConceptScore W4385474536C114614502 @default.
- W4385474536 hasConceptScore W4385474536C12267149 @default.
- W4385474536 hasConceptScore W4385474536C126255220 @default.
- W4385474536 hasConceptScore W4385474536C127162648 @default.
- W4385474536 hasConceptScore W4385474536C134306372 @default.
- W4385474536 hasConceptScore W4385474536C14036430 @default.
- W4385474536 hasConceptScore W4385474536C149728462 @default.
- W4385474536 hasConceptScore W4385474536C154945302 @default.
- W4385474536 hasConceptScore W4385474536C162324750 @default.
- W4385474536 hasConceptScore W4385474536C185429906 @default.
- W4385474536 hasConceptScore W4385474536C202444582 @default.
- W4385474536 hasConceptScore W4385474536C22324862 @default.
- W4385474536 hasConceptScore W4385474536C2777303404 @default.
- W4385474536 hasConceptScore W4385474536C28826006 @default.
- W4385474536 hasConceptScore W4385474536C31258907 @default.
- W4385474536 hasConceptScore W4385474536C33676613 @default.
- W4385474536 hasConceptScore W4385474536C33923547 @default.
- W4385474536 hasConceptScore W4385474536C39927690 @default.
- W4385474536 hasConceptScore W4385474536C41008148 @default.
- W4385474536 hasConceptScore W4385474536C50522688 @default.
- W4385474536 hasConceptScore W4385474536C50644808 @default.
- W4385474536 hasConceptScore W4385474536C57869625 @default.
- W4385474536 hasConceptScore W4385474536C66905080 @default.
- W4385474536 hasConceptScore W4385474536C78458016 @default.
- W4385474536 hasConceptScore W4385474536C81388566 @default.
- W4385474536 hasConceptScore W4385474536C86803240 @default.
- W4385474536 hasLocation W43854745361 @default.
- W4385474536 hasOpenAccess W4385474536 @default.
- W4385474536 hasPrimaryLocation W43854745361 @default.
- W4385474536 hasRelatedWork W1988224349 @default.
- W4385474536 hasRelatedWork W1993095755 @default.
- W4385474536 hasRelatedWork W2417978703 @default.
- W4385474536 hasRelatedWork W2618253902 @default.
- W4385474536 hasRelatedWork W2726819869 @default.
- W4385474536 hasRelatedWork W3026892677 @default.
- W4385474536 hasRelatedWork W3123935509 @default.
- W4385474536 hasRelatedWork W4286857060 @default.
- W4385474536 hasRelatedWork W4299302254 @default.
- W4385474536 hasRelatedWork W4361762909 @default.
- W4385474536 isParatext "false" @default.
- W4385474536 isRetracted "false" @default.
- W4385474536 workType "article" @default.