FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3211226311> ?p ?o ?g. }

• W3211226311 abstract "The two-tower architecture has been widely applied for learning item and user representations, which is important for large-scale recommender systems. Many two-tower models are trained using various in-batch negative sampling strategies, where the effects of such strategies inherently rely on the size of mini-batches. However, training two-tower models with a large batch size is inefficient, as it demands a large volume of memory for item and user contents and consumes a lot of time for feature encoding. Interestingly, we find that neural encoders can output relatively stable features for the same input after warming up in the training process. Based on such facts, we propose a simple yet effective sampling strategy called Cross-Batch Negative Sampling (CBNS), which takes advantage of the encoded item embeddings from recent mini-batches to boost the model training. Both theoretical analysis and empirical evaluations demonstrate the effectiveness and the efficiency of CBNS." @default.
• W3211226311 created "2021-11-08" @default.
• W3211226311 creator A5035927942 @default.
• W3211226311 creator A5048669373 @default.
• W3211226311 creator A5083350101 @default.
• W3211226311 date "2021-10-28" @default.
• W3211226311 modified "2023-09-27" @default.
• W3211226311 title "Cross-Batch Negative Sampling for Training Two-Tower Recommenders" @default.
• W3211226311 cites W1558797106 @default.
• W3211226311 cites W1610356397 @default.
• W3211226311 cites W2027731328 @default.
• W3211226311 cites W2101409192 @default.
• W3211226311 cites W21207210 @default.
• W3211226311 cites W2152808281 @default.
• W3211226311 cites W2153579005 @default.
• W3211226311 cites W2158515176 @default.
• W3211226311 cites W2262817822 @default.
• W3211226311 cites W2295739661 @default.
• W3211226311 cites W2512971201 @default.
• W3211226311 cites W2640408555 @default.
• W3211226311 cites W2648699835 @default.
• W3211226311 cites W2741249238 @default.
• W3211226311 cites W2798916557 @default.
• W3211226311 cites W2912967843 @default.
• W3211226311 cites W2936133855 @default.
• W3211226311 cites W2963085847 @default.
• W3211226311 cites W2963350250 @default.
• W3211226311 cites W2964121744 @default.
• W3211226311 cites W2964324019 @default.
• W3211226311 cites W2972801466 @default.
• W3211226311 cites W2982108874 @default.
• W3211226311 cites W2982902390 @default.
• W3211226311 cites W2987249037 @default.
• W3211226311 cites W3023045848 @default.
• W3211226311 cites W3029062980 @default.
• W3211226311 cites W3035014997 @default.
• W3211226311 cites W3035524453 @default.
• W3211226311 cites W3038033387 @default.
• W3211226311 cites W3038572442 @default.
• W3211226311 cites W3080642298 @default.
• W3211226311 cites W3092683697 @default.
• W3211226311 cites W3098649723 @default.
• W3211226311 cites W3099700870 @default.
• W3211226311 cites W3103931423 @default.
• W3211226311 cites W3104748221 @default.
• W3211226311 cites W3118668786 @default.
• W3211226311 doi "https://doi.org/10.48550/arxiv.2110.15154" @default.
• W3211226311 hasPublicationYear "2021" @default.
• W3211226311 type Work @default.
• W3211226311 sameAs 3211226311 @default.
• W3211226311 citedByCount "0" @default.
• W3211226311 crossrefType "posted-content" @default.
• W3211226311 hasAuthorship W3211226311A5035927942 @default.
• W3211226311 hasAuthorship W3211226311A5048669373 @default.
• W3211226311 hasAuthorship W3211226311A5083350101 @default.
• W3211226311 hasBestOaLocation W32112263111 @default.
• W3211226311 hasConcept C106131492 @default.
• W3211226311 hasConcept C111919701 @default.
• W3211226311 hasConcept C118505674 @default.
• W3211226311 hasConcept C119857082 @default.
• W3211226311 hasConcept C121332964 @default.
• W3211226311 hasConcept C124101348 @default.
• W3211226311 hasConcept C125411270 @default.
• W3211226311 hasConcept C127413603 @default.
• W3211226311 hasConcept C138885662 @default.
• W3211226311 hasConcept C140779682 @default.
• W3211226311 hasConcept C147176958 @default.
• W3211226311 hasConcept C153294291 @default.
• W3211226311 hasConcept C154945302 @default.
• W3211226311 hasConcept C2776401178 @default.
• W3211226311 hasConcept C2777211547 @default.
• W3211226311 hasConcept C2777831296 @default.
• W3211226311 hasConcept C31972630 @default.
• W3211226311 hasConcept C41008148 @default.
• W3211226311 hasConcept C41895202 @default.
• W3211226311 hasConcept C50644808 @default.
• W3211226311 hasConcept C98045186 @default.
• W3211226311 hasConceptScore W3211226311C106131492 @default.
• W3211226311 hasConceptScore W3211226311C111919701 @default.
• W3211226311 hasConceptScore W3211226311C118505674 @default.
• W3211226311 hasConceptScore W3211226311C119857082 @default.
• W3211226311 hasConceptScore W3211226311C121332964 @default.
• W3211226311 hasConceptScore W3211226311C124101348 @default.
• W3211226311 hasConceptScore W3211226311C125411270 @default.
• W3211226311 hasConceptScore W3211226311C127413603 @default.
• W3211226311 hasConceptScore W3211226311C138885662 @default.
• W3211226311 hasConceptScore W3211226311C140779682 @default.
• W3211226311 hasConceptScore W3211226311C147176958 @default.
• W3211226311 hasConceptScore W3211226311C153294291 @default.
• W3211226311 hasConceptScore W3211226311C154945302 @default.
• W3211226311 hasConceptScore W3211226311C2776401178 @default.
• W3211226311 hasConceptScore W3211226311C2777211547 @default.
• W3211226311 hasConceptScore W3211226311C2777831296 @default.
• W3211226311 hasConceptScore W3211226311C31972630 @default.
• W3211226311 hasConceptScore W3211226311C41008148 @default.
• W3211226311 hasConceptScore W3211226311C41895202 @default.
• W3211226311 hasConceptScore W3211226311C50644808 @default.
• W3211226311 hasConceptScore W3211226311C98045186 @default.
• W3211226311 hasLocation W32112263111 @default.
• W3211226311 hasOpenAccess W3211226311 @default.

• next