知识卡片

普通读书笔记卡 · 8939.f · 生成模型:GPT-6

状态适应不同于参数更新

理解等级:中级

递归模型在输入到来时更新状态,可随上下文改变行为,即使模型权重保持不变。它与在线梯度训练更新参数是两种适应方式。反馈提供调整机制,但运行更久也可能积累误差或遗忘;必须用任务表现检验,不能从有反馈直接推出能力必然随时间提高。

这张卡片看不懂?

已有反馈 · 0 条

    还没有人提交反馈。

    在整书笔记中的位置

    原书出处 · 1 处

    • · 查看原书依据

      Recurrent connections give neural nets the ability to maintain a persistent state representation, which can be propagated forward in time or can be updated in the face of incoming experience. The same idea has been proposed in multiple fields: in control theory it is the basic distinction between open-loop and closed-loop control. Closed-loop control allows the controller to change its behavior based on feedback from

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第25章《循环神经网络》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)/_epub-src/OPS/xhtml/c025.xhtml) - 结论依据:摘录中关于状态适应不同于参数更新的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。 - 原始内容:Recurrent connections give neural nets the ability to maintain a persistent state representation, which can be propagated forward in time or can be updated in the face of incoming experience. The same idea has been proposed in multiple fields: in control theory it is the basic distinction between open-loop and closed-loop control. Closed-loop control allows the controller to change its behavior based on feedback from