Сравнение методов

	Двунаправленная РНС	Механизм внимания	Многоголовочное самовнимание
Область	Глубокое обучение	Глубокое обучение	Глубокое обучение
Семейство	Machine learning	Machine learning	Machine learning
Год появления	1997	2015	2017
Автор метода	Schuster, M. & Paliwal, K.K.	Bahdanau, D.; Luong, M.T.	Vaswani, A. et al.
Тип	Recurrent neural network (sequence model)	Neural attention layer (encoder-decoder)	Attention mechanism (Transformer core)
Основополагающий источник	Schuster, M. & Paliwal, K.K. (1997). Bidirectional Recurrent Neural Networks. IEEE Transactions on Signal Processing, 45(11), 2673–2681.	Bahdanau, D., Cho, K. & Bengio, Y. (2015). Neural Machine Translation by Jointly Learning to Align and Translate. ICLR.	Vaswani, A. et al. (2017). Attention Is All You Need. NeurIPS.
Другие названия	Çift Yönlü RNN / BiLSTM / BiGRU, bidirectional recurrent neural network, BiLSTM, BiGRU	Dikkat Mekanizması (Bahdanau / Luong Attention), dikkat mekanizmasi, neural attention, additive attention	Öz-Dikkat ve Çok Başlı Dikkat (Multi-Head Self-Attention), öz-dikkat, multi-head attention, scaled dot-product attention
Связанные	5	5	5
Сводка	A Bidirectional RNN, introduced by Schuster and Paliwal in 1997, processes a sequence in both forward and backward directions so that every position has access to its full surrounding context. With LSTM or GRU cells (BiLSTM/BiGRU) it is the standard approach for named-entity recognition, sequence labelling, and speech recognition.	The attention mechanism, introduced by Bahdanau, Cho and Bengio in 2015 and refined by Luong, Pham and Manning the same year, lets a sequence decoder dynamically learn which of the encoder's outputs to focus on at each step. Before the Transformer, it substantially improved machine-translation quality by freeing models from compressing an entire input into a single fixed vector.	Multi-head self-attention, introduced by Vaswani and colleagues in 2017, is the mechanism that lets every position in a sequence compute its relationship to all other positions in parallel. It is the core of the Transformer architecture and the foundation underneath BERT, GPT, and T5.