更改

* '''自注意力层简化表达'''：

* '''自注意力层简化表达'''：

<math>[math]\displaystyle{  

<math>\displaystyle{  

x_{t+1} = \sigma(Rx_t) \cdot \frac{\exp(Kx_t) + \exp(W) \cdot b_t}{\exp(Kx_t) \cdot (Vx_t) + \exp(W) \cdot a_t}  

x_{t+1} = \sigma(Rx_t) \cdot \frac{\exp(Kx_t) + \exp(W) \cdot a_t}{\exp(Kx_t) \cdot (Vx_t) + \exp(W) \cdot b_t}  

}[/math]</math>

}</math>

----

----