Matches in SemOpenAlex for { <https://semopenalex.org/work/W3204564014> ?p ?o ?g. }
Showing items 1 to 78 of
78
with 100 items per page.
- W3204564014 endingPage "853" @default.
- W3204564014 startingPage "844" @default.
- W3204564014 abstract "We consider class incremental learning (CIL) problem, in which a learning agent continuously learns new classes from incrementally arriving training data batches and aims to predict well on all the classes learned so far. The main challenge of the problem is the catastrophic forgetting, and for the exemplar-memory based CIL methods, it is generally known that the forgetting is commonly caused by the prediction score bias that is injected due to the data imbalance between the new classes and the old classes (in the exemplar-memory). While several methods have been proposed to correct such score bias by some additional post-processing, e.g., score re-scaling or balanced fine-tuning, no systematic analysis on the root cause of such bias has been done. To that end, we analyze that computing the softmax probabilities by combining the output scores for all old and new classes could be the main source of the bias and propose a new CIL method, Separated Softmax for Incremental Learning (SS-IL). Our SS-IL consists of separated softmax (SS) output layer and ratio-preserving (RP) mini-batches combined with task-wise knowledge distillation (TKD), and through extensive experimental results, we show our SS-IL achieves very strong state-of-the-art accuracy on several large-scale benchmarks. We also show SS-IL makes much more balanced prediction, without any additional post-processing steps as is done in other baselines." @default.
- W3204564014 created "2021-10-11" @default.
- W3204564014 creator A5006968060 @default.
- W3204564014 creator A5015964231 @default.
- W3204564014 creator A5024628582 @default.
- W3204564014 creator A5042649012 @default.
- W3204564014 creator A5046346613 @default.
- W3204564014 creator A5080346989 @default.
- W3204564014 date "2021-01-01" @default.
- W3204564014 modified "2023-09-23" @default.
- W3204564014 title "SS-IL: Separated Softmax for Incremental Learning" @default.
- W3204564014 hasPublicationYear "2021" @default.
- W3204564014 type Work @default.
- W3204564014 sameAs 3204564014 @default.
- W3204564014 citedByCount "1" @default.
- W3204564014 countsByYear W32045640142021 @default.
- W3204564014 crossrefType "proceedings-article" @default.
- W3204564014 hasAuthorship W3204564014A5006968060 @default.
- W3204564014 hasAuthorship W3204564014A5015964231 @default.
- W3204564014 hasAuthorship W3204564014A5024628582 @default.
- W3204564014 hasAuthorship W3204564014A5042649012 @default.
- W3204564014 hasAuthorship W3204564014A5046346613 @default.
- W3204564014 hasAuthorship W3204564014A5080346989 @default.
- W3204564014 hasConcept C108583219 @default.
- W3204564014 hasConcept C119857082 @default.
- W3204564014 hasConcept C127413603 @default.
- W3204564014 hasConcept C138885662 @default.
- W3204564014 hasConcept C153180895 @default.
- W3204564014 hasConcept C154945302 @default.
- W3204564014 hasConcept C188441871 @default.
- W3204564014 hasConcept C201995342 @default.
- W3204564014 hasConcept C2780451532 @default.
- W3204564014 hasConcept C2780735816 @default.
- W3204564014 hasConcept C41008148 @default.
- W3204564014 hasConcept C41895202 @default.
- W3204564014 hasConcept C7149132 @default.
- W3204564014 hasConceptScore W3204564014C108583219 @default.
- W3204564014 hasConceptScore W3204564014C119857082 @default.
- W3204564014 hasConceptScore W3204564014C127413603 @default.
- W3204564014 hasConceptScore W3204564014C138885662 @default.
- W3204564014 hasConceptScore W3204564014C153180895 @default.
- W3204564014 hasConceptScore W3204564014C154945302 @default.
- W3204564014 hasConceptScore W3204564014C188441871 @default.
- W3204564014 hasConceptScore W3204564014C201995342 @default.
- W3204564014 hasConceptScore W3204564014C2780451532 @default.
- W3204564014 hasConceptScore W3204564014C2780735816 @default.
- W3204564014 hasConceptScore W3204564014C41008148 @default.
- W3204564014 hasConceptScore W3204564014C41895202 @default.
- W3204564014 hasConceptScore W3204564014C7149132 @default.
- W3204564014 hasLocation W32045640141 @default.
- W3204564014 hasOpenAccess W3204564014 @default.
- W3204564014 hasPrimaryLocation W32045640141 @default.
- W3204564014 hasRelatedWork W1542298582 @default.
- W3204564014 hasRelatedWork W1873256004 @default.
- W3204564014 hasRelatedWork W1967128172 @default.
- W3204564014 hasRelatedWork W2106191340 @default.
- W3204564014 hasRelatedWork W2301276134 @default.
- W3204564014 hasRelatedWork W2623640756 @default.
- W3204564014 hasRelatedWork W2739553566 @default.
- W3204564014 hasRelatedWork W2913880062 @default.
- W3204564014 hasRelatedWork W2918756560 @default.
- W3204564014 hasRelatedWork W2995269283 @default.
- W3204564014 hasRelatedWork W3029225387 @default.
- W3204564014 hasRelatedWork W3108343974 @default.
- W3204564014 hasRelatedWork W3120093105 @default.
- W3204564014 hasRelatedWork W3154902308 @default.
- W3204564014 hasRelatedWork W3155850773 @default.
- W3204564014 hasRelatedWork W3177180004 @default.
- W3204564014 hasRelatedWork W3202959177 @default.
- W3204564014 hasRelatedWork W3207755942 @default.
- W3204564014 hasRelatedWork W3208484386 @default.
- W3204564014 hasRelatedWork W2243371756 @default.
- W3204564014 isParatext "false" @default.
- W3204564014 isRetracted "false" @default.
- W3204564014 magId "3204564014" @default.
- W3204564014 workType "article" @default.