Matches in SemOpenAlex for { <https://semopenalex.org/work/W4386721787> ?p ?o ?g. }
Showing items 1 to 59 of
59
with 100 items per page.
- W4386721787 abstract "Continuous unsupervised representation learning (CURL) research has greatly benefited from improvements in self-supervised learning (SSL) techniques. As a result, existing CURL methods using SSL can learn high-quality representations without any labels, but with a notable performance drop when learning on a many-tasks data stream. We hypothesize that this is caused by the regularization losses that are imposed to prevent forgetting, leading to a suboptimal plasticity-stability trade-off: they either do not adapt fully to the incoming data (low plasticity), or incur significant forgetting when allowed to fully adapt to a new SSL pretext-task (low stability). In this work, we propose to train an expert network that is relieved of the duty of keeping the previous knowledge and can focus on performing optimally on the new tasks (optimizing plasticity). In the second phase, we combine this new knowledge with the previous network in an adaptation-retrospection phase to avoid forgetting and initialize a new expert with the knowledge of the old network. We perform several experiments showing that our proposed approach outperforms other CURL exemplar-free methods in few- and many-task split settings. Furthermore, we show how to adapt our approach to semi-supervised continual learning (Semi-SCL) and show that we surpass the accuracy of other exemplar-free Semi-SCL methods and reach the results of some others that use exemplars." @default.
- W4386721787 created "2023-09-14" @default.
- W4386721787 creator A5017330235 @default.
- W4386721787 creator A5073531557 @default.
- W4386721787 creator A5081872084 @default.
- W4386721787 creator A5085356302 @default.
- W4386721787 date "2023-09-12" @default.
- W4386721787 modified "2023-09-30" @default.
- W4386721787 title "Plasticity-Optimized Complementary Networks for Unsupervised Continual Learning" @default.
- W4386721787 doi "https://doi.org/10.48550/arxiv.2309.06086" @default.
- W4386721787 hasPublicationYear "2023" @default.
- W4386721787 type Work @default.
- W4386721787 citedByCount "0" @default.
- W4386721787 crossrefType "posted-content" @default.
- W4386721787 hasAuthorship W4386721787A5017330235 @default.
- W4386721787 hasAuthorship W4386721787A5073531557 @default.
- W4386721787 hasAuthorship W4386721787A5081872084 @default.
- W4386721787 hasAuthorship W4386721787A5085356302 @default.
- W4386721787 hasBestOaLocation W43867217871 @default.
- W4386721787 hasConcept C119857082 @default.
- W4386721787 hasConcept C154945302 @default.
- W4386721787 hasConcept C15744967 @default.
- W4386721787 hasConcept C162324750 @default.
- W4386721787 hasConcept C180747234 @default.
- W4386721787 hasConcept C187736073 @default.
- W4386721787 hasConcept C2776135515 @default.
- W4386721787 hasConcept C2780451532 @default.
- W4386721787 hasConcept C41008148 @default.
- W4386721787 hasConcept C50644808 @default.
- W4386721787 hasConcept C7149132 @default.
- W4386721787 hasConcept C8038995 @default.
- W4386721787 hasConceptScore W4386721787C119857082 @default.
- W4386721787 hasConceptScore W4386721787C154945302 @default.
- W4386721787 hasConceptScore W4386721787C15744967 @default.
- W4386721787 hasConceptScore W4386721787C162324750 @default.
- W4386721787 hasConceptScore W4386721787C180747234 @default.
- W4386721787 hasConceptScore W4386721787C187736073 @default.
- W4386721787 hasConceptScore W4386721787C2776135515 @default.
- W4386721787 hasConceptScore W4386721787C2780451532 @default.
- W4386721787 hasConceptScore W4386721787C41008148 @default.
- W4386721787 hasConceptScore W4386721787C50644808 @default.
- W4386721787 hasConceptScore W4386721787C7149132 @default.
- W4386721787 hasConceptScore W4386721787C8038995 @default.
- W4386721787 hasLocation W43867217871 @default.
- W4386721787 hasOpenAccess W4386721787 @default.
- W4386721787 hasPrimaryLocation W43867217871 @default.
- W4386721787 hasRelatedWork W3007915134 @default.
- W4386721787 hasRelatedWork W3046775127 @default.
- W4386721787 hasRelatedWork W3123344745 @default.
- W4386721787 hasRelatedWork W3196155444 @default.
- W4386721787 hasRelatedWork W3208099188 @default.
- W4386721787 hasRelatedWork W3209574120 @default.
- W4386721787 hasRelatedWork W4285260836 @default.
- W4386721787 hasRelatedWork W4306321456 @default.
- W4386721787 hasRelatedWork W4367692580 @default.
- W4386721787 hasRelatedWork W4386462264 @default.
- W4386721787 isParatext "false" @default.
- W4386721787 isRetracted "false" @default.
- W4386721787 workType "article" @default.