该文档主要介绍了attention及其变种self attention 、multi-attention以及一些相关的paper

目录1. Seq2seq2. Transformer3. Self-Attention 机制详解4. Positional Encoding5. Layer Normalization6. Transformer Encoder 与 Decoder7. 总结Others 最近在家听贪心学院的NLP直播课。放到博客上作为NLP 课程的简单的梳理。 简介： ELMo等基于深度学习的方法可以有效地学习出上下文有关词向量，但毕竟是基于LSTM的序列模型，必然要面临梯度以及无法并行化的问题，即便结合使用注

Head-up displays (HUDs) enable a pilot to manage aircraft activities by facilitating simultaneous access to the flight instrument data and to the outside scene. However, HUDs can also distract a pilot. This study shows that HUD luminance non-uniformi