手法を比較

	マルチモーダルLSTM	アテンションメカニズム	Gated Recurrent Unit (GRU)	LSTM
分野	深層学習	深層学習	深層学習	深層学習
系統	Machine learning	Machine learning	Machine learning	Machine learning
提唱年	2016	2015	2014	1997
提唱者	Rajagopalan et al. and various concurrent works (2016–2018)	Bahdanau, D.; Luong, M.T.	Cho, K., van Merrienboer, B., Gulcehre, C., Bahdanau, D., Bougares, F., Schwenk, H., & Bengio, Y.	Hochreiter, S. & Schmidhuber, J.
種類	Recurrent neural network architecture	Neural attention layer (encoder-decoder)	Recurrent neural network with gating	Recurrent neural network (gated memory cell)
原典	Rajagopalan, S., Tran, L., Rozgic, V., Narayanan, S., Kumar, A., & Ramakrishna, S. (2016). Extending Long Short-Term Memory for Multi-View Structured Learning. In Proceedings of ECCV 2016. Springer.	Bahdanau, D., Cho, K. & Bengio, Y. (2015). Neural Machine Translation by Jointly Learning to Align and Translate. ICLR.	Cho, K., van Merrienboer, B., Gulcehre, C., Bahdanau, D., Bougares, F., Schwenk, H., & Bengio, Y. (2014). Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translation. In Proceedings of EMNLP 2014, pp. 1724–1734.	Hochreiter, S. & Schmidhuber, J. (1997). Long Short-Term Memory. Neural Computation, 9(8), 1735–1780.
別名	MM-LSTM, multimodal recurrent network, multi-input LSTM, multimodal sequence model	Dikkat Mekanizması (Bahdanau / Luong Attention), dikkat mekanizmasi, neural attention, additive attention	GRU, GRU network, gated RNN, GRU cell	LSTM (Uzun Kısa Dönem Bellek Ağı), long short-term memory, LSTM network, recurrent neural network with memory cells
関連	4	5	3	5
概要	Multimodal LSTM extends the standard Long Short-Term Memory network to jointly process sequential data from multiple input modalities — such as text, audio, and video — within a unified recurrent architecture. By fusing representations from different sources before or within the LSTM cells, it captures temporal dependencies that span and cross modalities, making it a foundational approach for tasks like sentiment analysis, video captioning, and affective computing.	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.	The Gated Recurrent Unit (GRU), introduced by Cho et al. in 2014, is a streamlined recurrent neural network that uses two learned gates — an update gate and a reset gate — to selectively retain or discard information across time steps, enabling effective sequence modelling with fewer parameters than LSTM.	LSTM (Long Short-Term Memory) is a recurrent neural network architecture, introduced by Sepp Hochreiter and Jürgen Schmidhuber in 1997, that can learn long-term dependencies in sequential data and is widely used for time-series and sequence prediction. It keeps an internal memory that lets information persist across many time steps.