知识卡片
二元交叉熵惩罚自信错误
对目标y和预测概率p,单样本损失为−[y log p+(1−y)log(1−p)]。y=1时p趋近0或y=0时p趋近1,损失增大;公式中的log(1−p)需有完整括号,概率输出与原始logit也不能混用。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
- · 神经网络基础
原书出处 · 1 处
- · 查看原书依据
Is this all you need to train your model? Not yet. There are two more crucial pieces your neuron needs to learn. One is the loss function, or cost function, you saw earlier in this chapter. The mean squared error (MSE), discussed in detail in appendix D, would be a good candidate for the error metric if this were a regression problem. For this problem,
参考来源
- 位置:《Natural Language Processing in Action, Second Edition》第05章《神经网络基础》(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-5.html)
- 结论依据:摘录中关于二元交叉熵惩罚自信错误的定义、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:Is this all you need to train your model? Not yet. There are two more crucial pieces your neuron needs to learn. One is the loss function, or cost function, you saw earlier in this chapter. The mean squared error (MSE), discussed in detail in appendix D, would be a good candidate for the error metric if this were a regression problem. For this problem,