知识卡片

普通读书笔记卡 · 6483.e · 生成模型:GPT-6

梯度提升拟合损失改进方向

理解等级:中级

梯度提升逐轮增加模型,使组合输出沿降低损失的方向改进。平方误差情形可用残差理解后续拟合目标,其他损失不一定等同原始标签减预测。每轮依赖之前的组合,因此树数与步长共同影响训练和过拟合。

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

      Gradient boosting focuses on residuals from earlier classifiers and fits the model to residuals. The gradient descent algorithm is employed to minimize the error. GBM applies boosted machine learning for emails extracted from each spam dataset. GBM constructs one tree at a time, where each new tree helps to rectify errors caused by the earlier trained tree, unlike the RF classification model, where the trees

    参考来源

    - 位置:《Artificial Intelligence and Data Science in Recommendation System: Current Trends, Technologies, and Applications》第02章《文本挖掘与垃圾邮件过滤》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Artificial Intelligence and Data Science in Recommendation System: Current Trends, Technologies, and Applications/_epub-src/OEBPS/Text/9789815136746-C2.html) - 结论依据:摘录中关于梯度提升拟合损失改进方向的定义、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:Gradient boosting focuses on residuals from earlier classifiers and fits the model to residuals. The gradient descent algorithm is employed to minimize the error. GBM applies boosted machine learning for emails extracted from each spam dataset. GBM constructs one tree at a time, where each new tree helps to rectify errors caused by the earlier trained tree, unlike the RF classification model, where the trees