Additive Attention Bahdanau et al., 2015 Uses a one layer feedforward network to calculate Attention Alignment Oh, basically it is the same as Bahdanau Attention