From 9bd398b90598931dfe741dc88eeb19431e1413aa Mon Sep 17 00:00:00 2001 From: Aston Zhang Date: Wed, 4 Jul 2018 01:01:58 +0000 Subject: [PATCH] update math --- chapter_natural-language-processing/attention.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/chapter_natural-language-processing/attention.md b/chapter_natural-language-processing/attention.md index e68a2d9..3d69f33 100644 --- a/chapter_natural-language-processing/attention.md +++ b/chapter_natural-language-processing/attention.md @@ -36,7 +36,7 @@ $$e_{t^\prime t} = \boldsymbol{v}^\top \tanh(\boldsymbol{W}_s \boldsymbol{s}_{t^ 其中$\boldsymbol{v}$、$\boldsymbol{W}_s$、$\boldsymbol{W}_h$以及编码器与解码器中的各个权重和偏差都是模型参数 [1]。 -Bahanau等在编码器和解码器中分别使用了门控循环单元 [1]。在解码器中，我们需要对门控循环单元的设计稍作修改。解码器在$t^\prime$时间步的隐藏状态为 +Bahanau等在编码器和解码器中分别使用了门控循环单元 [1]。在解码器中，我们需要对门控循环单元的设计稍作修改。解码器在$t^\prime $时间步的隐藏状态为 $$\boldsymbol{s}_{t^\prime} = \boldsymbol{z}_{t^\prime} \odot \boldsymbol{s}_{t^\prime-1} + (1 - \boldsymbol{z}_{t^\prime}) \odot \tilde{\boldsymbol{s}}_{t^\prime},$$ -- GitLab