Matches in SemOpenAlex for { <https://semopenalex.org/work/W4387427408> ?p ?o ?g. }
Showing items 1 to 67 of
67
with 100 items per page.
- W4387427408 abstract "Like generic multi-task learning, continual learning has the nature of multi-objective optimization, and therefore faces a trade-off between the performance of different tasks. That is, to optimize for the current task distribution, it may need to compromise performance on some previous tasks. This means that there exist multiple models that are Pareto-optimal at different times, each addressing a distinct task performance trade-off. Researchers have discussed how to train particular models to address specific trade-off preferences. However, existing algorithms require training overheads proportional to the number of preferences -- a large burden when there are multiple, possibly infinitely many, preferences. As a response, we propose Imprecise Bayesian Continual Learning (IBCL). Upon a new task, IBCL (1) updates a knowledge base in the form of a convex hull of model parameter distributions and (2) obtains particular models to address task trade-off preferences with zero-shot. That is, IBCL does not require any additional training overhead to generate preference-addressing models from its knowledge base. We show that models obtained by IBCL have guarantees in identifying the Pareto optimal parameters. Moreover, experiments on standard image classification and NLP tasks support this guarantee. Statistically, IBCL improves average per-task accuracy by at most 23% and peak per-task accuracy by at most 15% with respect to the baseline methods, with steadily near-zero or positive backward transfer. Most importantly, IBCL significantly reduces the training overhead from training 1 model per preference to at most 3 models for all preferences." @default.
- W4387427408 created "2023-10-08" @default.
- W4387427408 creator A5000098655 @default.
- W4387427408 creator A5020691490 @default.
- W4387427408 creator A5030456600 @default.
- W4387427408 creator A5081521997 @default.
- W4387427408 date "2023-10-04" @default.
- W4387427408 modified "2023-10-08" @default.
- W4387427408 title "IBCL: Zero-shot Model Generation for Task Trade-offs in Continual Learning" @default.
- W4387427408 doi "https://doi.org/10.48550/arxiv.2310.02995" @default.
- W4387427408 hasPublicationYear "2023" @default.
- W4387427408 type Work @default.
- W4387427408 citedByCount "0" @default.
- W4387427408 crossrefType "posted-content" @default.
- W4387427408 hasAuthorship W4387427408A5000098655 @default.
- W4387427408 hasAuthorship W4387427408A5020691490 @default.
- W4387427408 hasAuthorship W4387427408A5030456600 @default.
- W4387427408 hasAuthorship W4387427408A5081521997 @default.
- W4387427408 hasBestOaLocation W43874274081 @default.
- W4387427408 hasConcept C105795698 @default.
- W4387427408 hasConcept C111919701 @default.
- W4387427408 hasConcept C119857082 @default.
- W4387427408 hasConcept C126255220 @default.
- W4387427408 hasConcept C137635306 @default.
- W4387427408 hasConcept C138885662 @default.
- W4387427408 hasConcept C154945302 @default.
- W4387427408 hasConcept C162324750 @default.
- W4387427408 hasConcept C187736073 @default.
- W4387427408 hasConcept C2779960059 @default.
- W4387427408 hasConcept C2780451532 @default.
- W4387427408 hasConcept C2780813799 @default.
- W4387427408 hasConcept C2781249084 @default.
- W4387427408 hasConcept C33923547 @default.
- W4387427408 hasConcept C41008148 @default.
- W4387427408 hasConcept C41895202 @default.
- W4387427408 hasConceptScore W4387427408C105795698 @default.
- W4387427408 hasConceptScore W4387427408C111919701 @default.
- W4387427408 hasConceptScore W4387427408C119857082 @default.
- W4387427408 hasConceptScore W4387427408C126255220 @default.
- W4387427408 hasConceptScore W4387427408C137635306 @default.
- W4387427408 hasConceptScore W4387427408C138885662 @default.
- W4387427408 hasConceptScore W4387427408C154945302 @default.
- W4387427408 hasConceptScore W4387427408C162324750 @default.
- W4387427408 hasConceptScore W4387427408C187736073 @default.
- W4387427408 hasConceptScore W4387427408C2779960059 @default.
- W4387427408 hasConceptScore W4387427408C2780451532 @default.
- W4387427408 hasConceptScore W4387427408C2780813799 @default.
- W4387427408 hasConceptScore W4387427408C2781249084 @default.
- W4387427408 hasConceptScore W4387427408C33923547 @default.
- W4387427408 hasConceptScore W4387427408C41008148 @default.
- W4387427408 hasConceptScore W4387427408C41895202 @default.
- W4387427408 hasLocation W43874274081 @default.
- W4387427408 hasOpenAccess W4387427408 @default.
- W4387427408 hasPrimaryLocation W43874274081 @default.
- W4387427408 hasRelatedWork W2035018219 @default.
- W4387427408 hasRelatedWork W2090624569 @default.
- W4387427408 hasRelatedWork W2250701745 @default.
- W4387427408 hasRelatedWork W2477876258 @default.
- W4387427408 hasRelatedWork W2800717388 @default.
- W4387427408 hasRelatedWork W2953266770 @default.
- W4387427408 hasRelatedWork W2989283631 @default.
- W4387427408 hasRelatedWork W3013650182 @default.
- W4387427408 hasRelatedWork W4249605382 @default.
- W4387427408 hasRelatedWork W4313491656 @default.
- W4387427408 isParatext "false" @default.
- W4387427408 isRetracted "false" @default.
- W4387427408 workType "article" @default.