Matches in SemOpenAlex for { <https://semopenalex.org/work/W3084572912> ?p ?o ?g. }
- W3084572912 abstract "We derive a new margin-based regularization formulation, termed multi-margin regularization (MMR), for deep neural networks (DNNs). The MMR is inspired by principles that were applied in margin analysis of shallow linear classifiers, e.g., support vector machine (SVM). Unlike SVM, MMR is continuously scaled by the radius of the bounding sphere (i.e., the maximal norm of the feature vector in the data), which is constantly changing during training. We empirically demonstrate that by a simple supplement to the loss function, our method achieves better results on various classification tasks across domains. Using the same concept, we also derive a selective sampling scheme and demonstrate accelerated training of DNNs by selecting samples according to a minimal margin score (MMS). This score measures the minimal amount of displacement an input should undergo until its predicted classification is switched. We evaluate our proposed methods on three image classification tasks and six language text classification tasks. Specifically, we show improved empirical results on CIFAR10, CIFAR100 and ImageNet using state-of-the-art convolutional neural networks (CNNs) and BERT-BASE architecture for the MNLI, QQP, QNLI, MRPC, SST-2 and RTE benchmarks." @default.
- W3084572912 created "2020-09-21" @default.
- W3084572912 creator A5022450249 @default.
- W3084572912 creator A5044926866 @default.
- W3084572912 creator A5052008807 @default.
- W3084572912 date "2020-09-13" @default.
- W3084572912 modified "2023-09-27" @default.
- W3084572912 title "Margin-Based Regularization and Selective Sampling in Deep Neural Networks." @default.
- W3084572912 cites W131533222 @default.
- W3084572912 cites W1463091736 @default.
- W3084572912 cites W1497676395 @default.
- W3084572912 cites W1686810756 @default.
- W3084572912 cites W1714704734 @default.
- W3084572912 cites W1975846642 @default.
- W3084572912 cites W2108598243 @default.
- W3084572912 cites W2110630246 @default.
- W3084572912 cites W2119821739 @default.
- W3084572912 cites W2130158090 @default.
- W3084572912 cites W2148603752 @default.
- W3084572912 cites W2152808281 @default.
- W3084572912 cites W2168885649 @default.
- W3084572912 cites W2174940656 @default.
- W3084572912 cites W2177410802 @default.
- W3084572912 cites W2194775991 @default.
- W3084572912 cites W2251939518 @default.
- W3084572912 cites W2396767181 @default.
- W3084572912 cites W2401231614 @default.
- W3084572912 cites W2525778437 @default.
- W3084572912 cites W2529714286 @default.
- W3084572912 cites W2607892599 @default.
- W3084572912 cites W2612445135 @default.
- W3084572912 cites W2622263826 @default.
- W3084572912 cites W2740018587 @default.
- W3084572912 cites W2746314669 @default.
- W3084572912 cites W2794302998 @default.
- W3084572912 cites W2804047946 @default.
- W3084572912 cites W2807007689 @default.
- W3084572912 cites W2911742574 @default.
- W3084572912 cites W2963207607 @default.
- W3084572912 cites W2963310665 @default.
- W3084572912 cites W2963341956 @default.
- W3084572912 cites W2963405349 @default.
- W3084572912 cites W2963685250 @default.
- W3084572912 cites W2964013229 @default.
- W3084572912 cites W2964153729 @default.
- W3084572912 cites W3017143921 @default.
- W3084572912 cites W3118608800 @default.
- W3084572912 cites W2963860801 @default.
- W3084572912 hasPublicationYear "2020" @default.
- W3084572912 type Work @default.
- W3084572912 sameAs 3084572912 @default.
- W3084572912 citedByCount "0" @default.
- W3084572912 crossrefType "posted-content" @default.
- W3084572912 hasAuthorship W3084572912A5022450249 @default.
- W3084572912 hasAuthorship W3084572912A5044926866 @default.
- W3084572912 hasAuthorship W3084572912A5052008807 @default.
- W3084572912 hasConcept C119857082 @default.
- W3084572912 hasConcept C12267149 @default.
- W3084572912 hasConcept C153180895 @default.
- W3084572912 hasConcept C154945302 @default.
- W3084572912 hasConcept C2776135515 @default.
- W3084572912 hasConcept C2984842247 @default.
- W3084572912 hasConcept C41008148 @default.
- W3084572912 hasConcept C50644808 @default.
- W3084572912 hasConcept C63584917 @default.
- W3084572912 hasConcept C774472 @default.
- W3084572912 hasConcept C81363708 @default.
- W3084572912 hasConcept C83665646 @default.
- W3084572912 hasConceptScore W3084572912C119857082 @default.
- W3084572912 hasConceptScore W3084572912C12267149 @default.
- W3084572912 hasConceptScore W3084572912C153180895 @default.
- W3084572912 hasConceptScore W3084572912C154945302 @default.
- W3084572912 hasConceptScore W3084572912C2776135515 @default.
- W3084572912 hasConceptScore W3084572912C2984842247 @default.
- W3084572912 hasConceptScore W3084572912C41008148 @default.
- W3084572912 hasConceptScore W3084572912C50644808 @default.
- W3084572912 hasConceptScore W3084572912C63584917 @default.
- W3084572912 hasConceptScore W3084572912C774472 @default.
- W3084572912 hasConceptScore W3084572912C81363708 @default.
- W3084572912 hasConceptScore W3084572912C83665646 @default.
- W3084572912 hasLocation W30845729121 @default.
- W3084572912 hasOpenAccess W3084572912 @default.
- W3084572912 hasPrimaryLocation W30845729121 @default.
- W3084572912 hasRelatedWork W1702573106 @default.
- W3084572912 hasRelatedWork W2002532360 @default.
- W3084572912 hasRelatedWork W2002844383 @default.
- W3084572912 hasRelatedWork W2016602863 @default.
- W3084572912 hasRelatedWork W2040790840 @default.
- W3084572912 hasRelatedWork W2067996809 @default.
- W3084572912 hasRelatedWork W2087539190 @default.
- W3084572912 hasRelatedWork W2091961126 @default.
- W3084572912 hasRelatedWork W2120287312 @default.
- W3084572912 hasRelatedWork W2149991085 @default.
- W3084572912 hasRelatedWork W2165155277 @default.
- W3084572912 hasRelatedWork W2171479471 @default.
- W3084572912 hasRelatedWork W2292974764 @default.
- W3084572912 hasRelatedWork W2551473848 @default.
- W3084572912 hasRelatedWork W2736123762 @default.
- W3084572912 hasRelatedWork W2827538589 @default.
- W3084572912 hasRelatedWork W2910396745 @default.