Matches in SemOpenAlex for { <https://semopenalex.org/work/W4379924928> ?p ?o ?g. }
Showing items 1 to 57 of
57
with 100 items per page.
- W4379924928 abstract "In classification problems with large output spaces (up to millions of labels), the last layer can require an enormous amount of memory. Using sparse connectivity would drastically reduce the memory requirements, but as we show below, it can result in much diminished predictive performance of the model. Fortunately, we found that this can be mitigated by introducing a penultimate layer of intermediate size. We further demonstrate that one can constrain the connectivity of the sparse layer to be uniform, in the sense that each output neuron will have the exact same number of incoming connections. This allows for efficient implementations of sparse matrix multiplication and connection redistribution on GPU hardware. Via a custom CUDA implementation, we show that the proposed approach can scale to datasets with 670,000 labels on a single commodity GPU with only 4GB memory." @default.
- W4379924928 created "2023-06-09" @default.
- W4379924928 creator A5017545333 @default.
- W4379924928 creator A5070761496 @default.
- W4379924928 date "2023-06-06" @default.
- W4379924928 modified "2023-09-23" @default.
- W4379924928 title "Towards Memory-Efficient Training for Extremely Large Output Spaces -- Learning with 500k Labels on a Single Commodity GPU" @default.
- W4379924928 doi "https://doi.org/10.48550/arxiv.2306.03725" @default.
- W4379924928 hasPublicationYear "2023" @default.
- W4379924928 type Work @default.
- W4379924928 citedByCount "0" @default.
- W4379924928 crossrefType "posted-content" @default.
- W4379924928 hasAuthorship W4379924928A5017545333 @default.
- W4379924928 hasAuthorship W4379924928A5070761496 @default.
- W4379924928 hasBestOaLocation W43799249281 @default.
- W4379924928 hasConcept C113775141 @default.
- W4379924928 hasConcept C114614502 @default.
- W4379924928 hasConcept C173608175 @default.
- W4379924928 hasConcept C178790620 @default.
- W4379924928 hasConcept C185592680 @default.
- W4379924928 hasConcept C199360897 @default.
- W4379924928 hasConcept C26713055 @default.
- W4379924928 hasConcept C2778119891 @default.
- W4379924928 hasConcept C2779227376 @default.
- W4379924928 hasConcept C2780595030 @default.
- W4379924928 hasConcept C33923547 @default.
- W4379924928 hasConcept C41008148 @default.
- W4379924928 hasConcept C9390403 @default.
- W4379924928 hasConceptScore W4379924928C113775141 @default.
- W4379924928 hasConceptScore W4379924928C114614502 @default.
- W4379924928 hasConceptScore W4379924928C173608175 @default.
- W4379924928 hasConceptScore W4379924928C178790620 @default.
- W4379924928 hasConceptScore W4379924928C185592680 @default.
- W4379924928 hasConceptScore W4379924928C199360897 @default.
- W4379924928 hasConceptScore W4379924928C26713055 @default.
- W4379924928 hasConceptScore W4379924928C2778119891 @default.
- W4379924928 hasConceptScore W4379924928C2779227376 @default.
- W4379924928 hasConceptScore W4379924928C2780595030 @default.
- W4379924928 hasConceptScore W4379924928C33923547 @default.
- W4379924928 hasConceptScore W4379924928C41008148 @default.
- W4379924928 hasConceptScore W4379924928C9390403 @default.
- W4379924928 hasLocation W43799249281 @default.
- W4379924928 hasOpenAccess W4379924928 @default.
- W4379924928 hasPrimaryLocation W43799249281 @default.
- W4379924928 hasRelatedWork W1507172387 @default.
- W4379924928 hasRelatedWork W2076165488 @default.
- W4379924928 hasRelatedWork W2090121768 @default.
- W4379924928 hasRelatedWork W2161462353 @default.
- W4379924928 hasRelatedWork W2295371547 @default.
- W4379924928 hasRelatedWork W2392023973 @default.
- W4379924928 hasRelatedWork W2558367073 @default.
- W4379924928 hasRelatedWork W2982613029 @default.
- W4379924928 hasRelatedWork W3013976982 @default.
- W4379924928 hasRelatedWork W3038415719 @default.
- W4379924928 isParatext "false" @default.
- W4379924928 isRetracted "false" @default.
- W4379924928 workType "article" @default.