FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3175149871> ?p ?o ?g. }

• W3175149871 abstract "The increasing computational requirements of deep neural networks (DNNs) have led to significant interest in obtaining DNN models that are sparse, yet accurate. Recent work has investigated the even harder case of sparse training, where the DNN weights are, for as much as possible, already sparse to reduce computational costs during training. Existing sparse training methods are often empirical and can have lower accuracy relative to the dense baseline. In this paper, we present a general approach called Alternating Compressed/DeCompressed (AC/DC) training of DNNs, demonstrate convergence for a variant of the algorithm, and show that AC/DC outperforms existing sparse training methods in accuracy at similar computational budgets; at high sparsity levels, AC/DC even outperforms existing methods that rely on accurate pre-trained dense models. An important property of AC/DC is that it allows co-training of dense and sparse models, yielding accurate sparse-dense model pairs at the end of the training process. This is useful in practice, where compressed variants may be desirable for deployment in resource-constrained settings without re-doing the entire training flow, and also provides us with insights into the accuracy gap between dense and compressed models. The code is available at: https://github.com/IST-DASLab/ACDC ." @default.
• W3175149871 created "2021-07-05" @default.
• W3175149871 creator A5000837557 @default.
• W3175149871 creator A5076938722 @default.
• W3175149871 creator A5083822059 @default.
• W3175149871 creator A5091392485 @default.
• W3175149871 date "2021-12-06" @default.
• W3175149871 modified "2023-10-18" @default.
• W3175149871 title "AC/DC: Alternating Compressed/DeCompressed Training of Deep Neural Networks" @default.
• W3175149871 cites W119997944 @default.
• W3175149871 cites W1522301498 @default.
• W3175149871 cites W1977520307 @default.
• W3175149871 cites W1980287119 @default.
• W3175149871 cites W1980454827 @default.
• W3175149871 cites W2037537409 @default.
• W3175149871 cites W2114766824 @default.
• W3175149871 cites W2117539524 @default.
• W3175149871 cites W2129638195 @default.
• W3175149871 cites W2156150815 @default.
• W3175149871 cites W2163224677 @default.
• W3175149871 cites W2194775991 @default.
• W3175149871 cites W2319920447 @default.
• W3175149871 cites W2499540656 @default.
• W3175149871 cites W2525332836 @default.
• W3175149871 cites W2582745083 @default.
• W3175149871 cites W2589646680 @default.
• W3175149871 cites W2612445135 @default.
• W3175149871 cites W2764043458 @default.
• W3175149871 cites W2768501777 @default.
• W3175149871 cites W2804032941 @default.
• W3175149871 cites W2915589364 @default.
• W3175149871 cites W2948635472 @default.
• W3175149871 cites W2956434358 @default.
• W3175149871 cites W2963247446 @default.
• W3175149871 cites W2963248893 @default.
• W3175149871 cites W2963403868 @default.
• W3175149871 cites W2963674932 @default.
• W3175149871 cites W2963765119 @default.
• W3175149871 cites W2963766684 @default.
• W3175149871 cites W2963813662 @default.
• W3175149871 cites W2964098911 @default.
• W3175149871 cites W2964110616 @default.
• W3175149871 cites W2990352720 @default.
• W3175149871 cites W2995435108 @default.
• W3175149871 cites W2995463996 @default.
• W3175149871 cites W3007957341 @default.
• W3175149871 cites W3034733718 @default.
• W3175149871 cites W3034848862 @default.
• W3175149871 cites W3035180000 @default.
• W3175149871 cites W3035304835 @default.
• W3175149871 cites W3037032032 @default.
• W3175149871 cites W3093348380 @default.
• W3175149871 cites W3098372854 @default.
• W3175149871 cites W3101584733 @default.
• W3175149871 cites W3104688113 @default.
• W3175149871 cites W3118608800 @default.
• W3175149871 cites W3127243802 @default.
• W3175149871 cites W3129093240 @default.
• W3175149871 cites W3137147200 @default.
• W3175149871 cites W3137695714 @default.
• W3175149871 cites W3156528192 @default.
• W3175149871 cites W3103414006 @default.
• W3175149871 hasPublicationYear "2021" @default.
• W3175149871 type Work @default.
• W3175149871 sameAs 3175149871 @default.
• W3175149871 citedByCount "3" @default.
• W3175149871 countsByYear W31751498712021 @default.
• W3175149871 crossrefType "proceedings-article" @default.
• W3175149871 hasAuthorship W3175149871A5000837557 @default.
• W3175149871 hasAuthorship W3175149871A5076938722 @default.
• W3175149871 hasAuthorship W3175149871A5083822059 @default.
• W3175149871 hasAuthorship W3175149871A5091392485 @default.
• W3175149871 hasBestOaLocation W31751498713 @default.
• W3175149871 hasConcept C111472728 @default.
• W3175149871 hasConcept C11413529 @default.
• W3175149871 hasConcept C119857082 @default.
• W3175149871 hasConcept C121332964 @default.
• W3175149871 hasConcept C124066611 @default.
• W3175149871 hasConcept C138885662 @default.
• W3175149871 hasConcept C153180895 @default.
• W3175149871 hasConcept C153294291 @default.
• W3175149871 hasConcept C154945302 @default.
• W3175149871 hasConcept C162324750 @default.
• W3175149871 hasConcept C177264268 @default.
• W3175149871 hasConcept C189950617 @default.
• W3175149871 hasConcept C199360897 @default.
• W3175149871 hasConcept C2776760102 @default.
• W3175149871 hasConcept C2777211547 @default.
• W3175149871 hasConcept C2777303404 @default.
• W3175149871 hasConcept C2984842247 @default.
• W3175149871 hasConcept C41008148 @default.
• W3175149871 hasConcept C50522688 @default.
• W3175149871 hasConcept C50644808 @default.
• W3175149871 hasConceptScore W3175149871C111472728 @default.
• W3175149871 hasConceptScore W3175149871C11413529 @default.
• W3175149871 hasConceptScore W3175149871C119857082 @default.
• W3175149871 hasConceptScore W3175149871C121332964 @default.
• W3175149871 hasConceptScore W3175149871C124066611 @default.
• W3175149871 hasConceptScore W3175149871C138885662 @default.
• W3175149871 hasConceptScore W3175149871C153180895 @default.

• next