Transcription of Co2L: Contrastive Continual Learning
{{id}} {{{paragraph}}}
Example: biology
Co2L: Contrastive Continual Learning
Co2L: Contrastive Continual LearningHyuntak ChaJaeho LeeJinwoo ShinKAISTD aejeon, South Korea{ , jaeho-lee, breakthroughs in self-supervised Learning showthat such algorithms learn visual representations that can betransferred better to unseen tasks than cross-entropy basedmethods which rely on task-specific supervision. In this pa-per, we found that the similar holds in the Continual learningcontext: contrastively learned representations are more ro-bust against the catastrophic forgetting than ones trainedwith the cross-entropy objective. Based on this novel ob-servation, we propose a rehearsal-based Continual learningalgorithm that focuses on continually Learning and main-taining transferable representations.}
continual learning literature focus on preserving the previ-ously learned knowledge using various types of information about the past task. Replay-based approaches store a small portion of past samples and rehearse the samples along with present task samples [35, 29, 34, 5]. Regularization-based
Loading..
Tags:
Information
Domain:
Source:
Link to this page:
Please notify us if you found a problem with this document: