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： Targeting cookies These cookies are takım on our site by our advertising partners. They may be used by relevant companies to build your interests profile and show relevant ads on other sites. 论文指出，门控网络倾向于收敛到一种状态，总是为相同的几个专家产生大的权重。这种不平衡是自我强化的... https://www.blogger.com/u/9/profile/10810167670069336925