Matches in SemOpenAlex for { <https://semopenalex.org/work/W2890599985> ?p ?o ?g. }
Showing items 1 to 99 of
99
with 100 items per page.
- W2890599985 abstract "Recently, path norm was proposed as a new capacity measure for neural networks with Rectified Linear Unit (ReLU) activation function, which takes the rescaling-invariant property of ReLU into account. It has been shown that the generalization error bound in terms of the path norm explains the empirical generalization behaviors of the ReLU neural networks better than that of other capacity measures. Moreover, optimization algorithms which take path norm as the regularization term to the loss function, like Path-SGD, have been shown to achieve better generalization performance. However, the path norm counts the values of all paths, and hence the capacity measure based on path norm could be improperly influenced by the dependency among different paths. It is also known that each path of a ReLU network can be represented by a small group of linearly independent basis paths with multiplication and division operation, which indicates that the generalization behavior of the network only depends on only a few basis paths. Motivated by this, we propose a new norm emph{Basis-path Norm} based on a group of linearly independent paths to measure the capacity of neural networks more accurately. We establish a generalization error bound based on this basis path norm, and show it explains the generalization behaviors of ReLU networks more accurately than previous capacity measures via extensive experiments. In addition, we develop optimization algorithms which minimize the empirical risk regularized by the basis-path norm. Our experiments on benchmark datasets demonstrate that the proposed regularization method achieves clearly better performance on the test set than the previous regularization approaches." @default.
- W2890599985 created "2018-09-27" @default.
- W2890599985 creator A5017541508 @default.
- W2890599985 creator A5044802273 @default.
- W2890599985 creator A5051093347 @default.
- W2890599985 creator A5064573190 @default.
- W2890599985 creator A5066928490 @default.
- W2890599985 creator A5070990160 @default.
- W2890599985 date "2018-09-19" @default.
- W2890599985 modified "2023-09-23" @default.
- W2890599985 title "Capacity Control of ReLU Neural Networks by Basis-path Norm" @default.
- W2890599985 cites W1904365287 @default.
- W2890599985 cites W2028988057 @default.
- W2890599985 cites W2110652811 @default.
- W2890599985 cites W2133564696 @default.
- W2890599985 cites W2194775991 @default.
- W2890599985 cites W2327501763 @default.
- W2890599985 cites W2401137308 @default.
- W2890599985 cites W2523246573 @default.
- W2890599985 cites W2533523411 @default.
- W2890599985 cites W2565948352 @default.
- W2890599985 cites W2579923771 @default.
- W2890599985 cites W2604117713 @default.
- W2890599985 cites W2613904329 @default.
- W2890599985 cites W2626778328 @default.
- W2890599985 cites W2709553318 @default.
- W2890599985 cites W2766196653 @default.
- W2890599985 cites W2949117887 @default.
- W2890599985 cites W2951581544 @default.
- W2890599985 cites W2952062734 @default.
- W2890599985 cites W2953106684 @default.
- W2890599985 cites W2963446712 @default.
- W2890599985 cites W2963695615 @default.
- W2890599985 cites W2963739978 @default.
- W2890599985 cites W3118608800 @default.
- W2890599985 cites W577198184 @default.
- W2890599985 doi "https://doi.org/10.48550/arxiv.1809.07122" @default.
- W2890599985 hasPublicationYear "2018" @default.
- W2890599985 type Work @default.
- W2890599985 sameAs 2890599985 @default.
- W2890599985 citedByCount "4" @default.
- W2890599985 countsByYear W28905999852018 @default.
- W2890599985 countsByYear W28905999852019 @default.
- W2890599985 countsByYear W28905999852021 @default.
- W2890599985 crossrefType "posted-content" @default.
- W2890599985 hasAuthorship W2890599985A5017541508 @default.
- W2890599985 hasAuthorship W2890599985A5044802273 @default.
- W2890599985 hasAuthorship W2890599985A5051093347 @default.
- W2890599985 hasAuthorship W2890599985A5064573190 @default.
- W2890599985 hasAuthorship W2890599985A5066928490 @default.
- W2890599985 hasAuthorship W2890599985A5070990160 @default.
- W2890599985 hasBestOaLocation W28905999851 @default.
- W2890599985 hasConcept C11413529 @default.
- W2890599985 hasConcept C126255220 @default.
- W2890599985 hasConcept C134306372 @default.
- W2890599985 hasConcept C154945302 @default.
- W2890599985 hasConcept C177148314 @default.
- W2890599985 hasConcept C17744445 @default.
- W2890599985 hasConcept C191795146 @default.
- W2890599985 hasConcept C199360897 @default.
- W2890599985 hasConcept C199539241 @default.
- W2890599985 hasConcept C2777735758 @default.
- W2890599985 hasConcept C33923547 @default.
- W2890599985 hasConcept C38365724 @default.
- W2890599985 hasConcept C41008148 @default.
- W2890599985 hasConcept C50644808 @default.
- W2890599985 hasConcept C5917680 @default.
- W2890599985 hasConceptScore W2890599985C11413529 @default.
- W2890599985 hasConceptScore W2890599985C126255220 @default.
- W2890599985 hasConceptScore W2890599985C134306372 @default.
- W2890599985 hasConceptScore W2890599985C154945302 @default.
- W2890599985 hasConceptScore W2890599985C177148314 @default.
- W2890599985 hasConceptScore W2890599985C17744445 @default.
- W2890599985 hasConceptScore W2890599985C191795146 @default.
- W2890599985 hasConceptScore W2890599985C199360897 @default.
- W2890599985 hasConceptScore W2890599985C199539241 @default.
- W2890599985 hasConceptScore W2890599985C2777735758 @default.
- W2890599985 hasConceptScore W2890599985C33923547 @default.
- W2890599985 hasConceptScore W2890599985C38365724 @default.
- W2890599985 hasConceptScore W2890599985C41008148 @default.
- W2890599985 hasConceptScore W2890599985C50644808 @default.
- W2890599985 hasConceptScore W2890599985C5917680 @default.
- W2890599985 hasLocation W28905999851 @default.
- W2890599985 hasOpenAccess W2890599985 @default.
- W2890599985 hasPrimaryLocation W28905999851 @default.
- W2890599985 hasRelatedWork W1971753667 @default.
- W2890599985 hasRelatedWork W1994818536 @default.
- W2890599985 hasRelatedWork W1997271576 @default.
- W2890599985 hasRelatedWork W2052275678 @default.
- W2890599985 hasRelatedWork W2068208308 @default.
- W2890599985 hasRelatedWork W2348187822 @default.
- W2890599985 hasRelatedWork W2536463469 @default.
- W2890599985 hasRelatedWork W2792268997 @default.
- W2890599985 hasRelatedWork W2998088892 @default.
- W2890599985 hasRelatedWork W3012292080 @default.
- W2890599985 isParatext "false" @default.
- W2890599985 isRetracted "false" @default.
- W2890599985 magId "2890599985" @default.
- W2890599985 workType "article" @default.