Comparer des méthodes

	Longformer / BigBird	Mélange d'experts	XGBoost
Domaine	Apprentissage profond	Apprentissage profond	Apprentissage automatique
Famille	Machine learning	Machine learning	Machine learning
Année d'origine	2020	2017	2016
Auteur d'origine	Beltagy, Peters & Cohan (Longformer); Zaheer et al. (BigBird)	Shazeer, N. et al.	Chen, T. & Guestrin, C.
Type	Sparse-attention Transformer for long sequences	Sparse neural network architecture (conditional computation)	Ensemble (gradient-boosted decision trees)
Source fondatrice	Beltagy, I., Peters, M. E. & Cohan, A. (2020). Longformer: The Long-Document Transformer. arXiv.	Shazeer, N. et al. (2017). Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer. ICLR. arXiv:1701.06538	Chen, T. & Guestrin, C. (2016). XGBoost: A Scalable Tree Boosting System. Proceedings of the 22nd ACM SIGKDD, 785–794.
Alias	Uzun Dizi Transformer (Longformer / BigBird), uzun dizi transformer, long-document transformer, sparse-attention transformer	Uzman Karışımı (Mixture of Experts — MoE), uzman karışımı, MoE, sparse mixture of experts	XGBoost, extreme gradient boosting, scalable tree boosting
Apparentées	4	3	5
Résumé	Long-sequence Transformers such as Longformer (Beltagy, Peters & Cohan, 2020) and BigBird (Zaheer et al., 2020) replace the standard Transformer's O(n²) attention with sparse attention patterns that scale linearly, O(n), with sequence length. This lets a single model attend over thousands of tokens — full documents, legal texts, or genomic sequences — that would not fit a conventional Transformer.	Mixture of Experts (MoE) is a sparse neural-network architecture, introduced by Shazeer and colleagues in 2017 with the sparsely-gated MoE layer, in which only a subset of expert sub-networks is activated for each input. As seen in models such as Switch Transformer and Mixtral, it holds computation cost fixed even as the total parameter count grows.	XGBoost (Extreme Gradient Boosting) is a scalable tree-boosting algorithm introduced by Tianqi Chen and Carlos Guestrin in 2016. It builds a strong predictor by adding decision trees one at a time, each correcting the errors left by the trees before it, and is a powerful prediction method widely used in competitions.