Compară metode

	Longformer / BigBird	Rețeaua de Atenție Grafică	Amestec de Experți
Domeniu	Învățare profundă	Învățare profundă	Învățare profundă
Familie	Machine learning	Machine learning	Machine learning
Anul apariției	2020	2018	2017
Autorul original	Beltagy, Peters & Cohan (Longformer); Zaheer et al. (BigBird)	Veličković, P. et al.	Shazeer, N. et al.
Tip	Sparse-attention Transformer for long sequences	Graph neural network (attention-based)	Sparse neural network architecture (conditional computation)
Sursa seminală	Beltagy, I., Peters, M. E. & Cohan, A. (2020). Longformer: The Long-Document Transformer. arXiv.	Veličković, P. et al. (2018). Graph Attention Networks. ICLR.	Shazeer, N. et al. (2017). Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer. ICLR. arXiv:1701.06538
Denumiri alternative	Uzun Dizi Transformer (Longformer / BigBird), uzun dizi transformer, long-document transformer, sparse-attention transformer	Graf Dikkat Ağı (GAT), GAT, graph attention network, attention-based graph neural network	Uzman Karışımı (Mixture of Experts — MoE), uzman karışımı, MoE, sparse mixture of experts
Înrudite	4	4	3
Rezumat	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.	The Graph Attention Network (GAT), introduced by Veličković and colleagues in 2018, is a graph neural network variant that learns how much importance to assign to each neighbouring node through a self-attention mechanism. On heterogeneous neighbourhoods and relational classification it produces results superior to graph convolutional networks (GCN).	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.