← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 6699.a · 生成模型:GPT-6

概率示例应同时检查数值与总和

理解等级:中级

向量(0.5,0.9,0.2)经softmax约为(0.309,0.461,0.229)。原图第三项0.929使总和超过1,属于数值错误;小例手算可核对公式、顺序与解释是否一致,不把错误图值作为算法行为。

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

      The output vector of a three-neuron output layer will look something like this 3D column vector: Equation 6.4 Example 3D vector Then, the “squashed” vector after the softmax activation would look like equation 6.5. Notice how the rank order of the values in the vector have not changed, but their sum now equals 1. Equation 6.5 Example 3D vector after softmax Notice that the sum

    参考来源

    - 位置:《Natural Language Processing in Action, Second Edition》第06章《用词嵌入推理》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Natural Language Processing in Action, Second Edition/_epub-src/OEBPS/Text/chapter-6.html) - 结论依据:摘录中关于概率示例应同时检查数值与总和的定义、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:The output vector of a three-neuron output layer will look something like this 3D column vector: Equation 6.4 Example 3D vector Then, the “squashed” vector after the softmax activation would look like equation 6.5. Notice how the rank order of the values in the vector have not changed, but their sum now equals 1. Equation 6.5 Example 3D vector after softmax Notice that the sum