设 和 分别是门控网络和第 个 Expert 的输出，那么对于在当前的输入x，输出就是所有 experts 的加权和： They work by uniquely identifying your browser and device. If you do derece allow these cookies, we could derece be able to offer you a specialized ad experience on different sites. 对比一下可以看出，在计算每个 expert 的损失之后，先把它给指数化了再进行加权求和，最后取了log。这也是一个我们... https://www.blogger.com/u/9/profile/10810167670069336925