@prefix owl: <http://www.w3.org/2002/07/owl#> .
@prefix skos: <http://www.w3.org/2004/02/skos/core#> .
@prefix dc: <http://purl.org/dc/terms/> .
@prefix xsd: <http://www.w3.org/2001/XMLSchema#> .

<http://data.loterre.fr/ark:/67375/8LP> a owl:Ontology, skos:ConceptScheme .
<http://data.loterre.fr/ark:/67375/8LP-CQ937GB8-K>
  skos:prefLabel "attention mechanism"@en, "mécanisme d'attention"@fr ;
  a skos:Concept ;
  skos:narrower <http://data.loterre.fr/ark:/67375/8LP-C01T3CNT-T> .

<http://data.loterre.fr/ark:/67375/8LP-C01T3CNT-T>
  skos:broader <http://data.loterre.fr/ark:/67375/8LP-CQ937GB8-K> ;
  a skos:Concept ;
  skos:inScheme <http://data.loterre.fr/ark:/67375/8LP> ;
  skos:prefLabel "distribution d'attention"@fr, "attention distribution"@en ;
  skos:example "Recent research in language processing finds that attention weights are not a good proxy for relative importance because different attention distributions can lead to the same predictions (Jain and Wallace 2019). (Hollenstein & Beinborn, 2021)"@en, "As a result the ideal attention distribution should put all of the probability mass on the antecedent noun phrase for reflexive anaphora or on the sub-ject noun phrase for agreement and zero on the distractor noun phrases. (Lin, Tan & Frank, 2019)"@en, "Recent research indicates that complementary attention distributions can lead to the same model prediction (Jain and Wallace 2019; Wiegreffe and Pinter 2019) and that the removal of input tokens with large attention weights often does not lead to a change in the model's prediction (Serrano and Smith 2019). (Hollenstein & Beinborn, 2021)"@en, "Further analysis indicates that WID can also learn the attention patterns from the teacher model without any alignment loss on attention distributions. (Wu, Hou, Lao, Li, Wong, Zhao & Yang, 2024)"@en, "The proposed method aims to unravel the attention distribution at each layer within a multi-layer model. (Jang, Byun & Shin, 2024)"@en ;
  skos:definition "The distribution of attention weights across the input sequence in an attention-based model."@en, "Manière dont les poids d'attention sont répartis dans la séquence d'entrée d'un modèle basé sur l'attention."@fr ;
  skos:hiddenLabel "Attention distribution"@en, "Distribution d'attention"@fr ;
  dc:modified "2024-05-27T07:33:05"^^xsd:dateTime .