知识卡片

普通读书笔记卡 · 9114.b · 生成模型:GPT-6

参数与数据在计算图中职责不同

理解等级:初级

参数θ包含权重、偏置和其它可学习量;数据x既可指原始输入,也可指隐藏激活或输出。讨论特定层时用输入与输出标记能避免层号混淆;若区分激活前后,需明确z与h等约定。不要因为都是张量,就忽略哪些量在训练中被更新。

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

      Parameters: θ. These include weights, W, and biases, b, as well as any other learnable parameters of a network. Data: x. Data can refer to inputs to the network, activations on hidden layers, outputs from the network, and so on. Any representation of the signal being processed is considered to be data. Sometimes we will wish to distinguish between the raw inputs, hidden units, and

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》符号约定专题《符号与矩阵微积分约定》(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/c000b.xhtml) - 结论依据:摘录中关于参数与数据在计算图中职责不同的定义、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:Parameters: θ. These include weights, W, and biases, b, as well as any other learnable parameters of a network. Data: x. Data can refer to inputs to the network, activations on hidden layers, outputs from the network, and so on. Any representation of the signal being processed is considered to be data. Sometimes we will wish to distinguish between the raw inputs, hidden units, and