In 2017, Google Stopped Fixing the Flaw. They Removed It.
中文摘要
谷歌利用注意力机制架构,彻底解决了梯度消失问题
English Summary
In 2017, Google solved the vanishing gradient problem by introducing the attention mechanism architecture, effectively removing the flaw instead of just fixing it.
原文节选
How Attention Works: The Architecture That Broke the Vanishing Gradient Continue reading on Towards AI »