知识卡片

普通读书笔记卡 · 6484.h · 生成模型:GPT-6

折扣增益重视前排相关性

理解等级:中级

DCG把每个结果的相关增益按位置折扣,较靠前的相关结果贡献更大;NDCG再除以同一候选相关性下的理想排序得分。比较时要统一增益、折扣与截断长度。归一化提供可比尺度,不意味着所有任务和候选难度完全可比。

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

      Discounted Cumulative Gain (DCG) is the metric that rates and evaluates the gain of each item based on its position in the recommended list of an RS. DCG assumes the earlier arriving items to have more preference. Discounting based on the item position uses a logarithmic factor, as shown below.

    参考来源

    - 位置:《Artificial Intelligence and Data Science in Recommendation System: Current Trends, Technologies, and Applications》第03章《深度推荐与评估指标》(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-C3.html) - 结论依据:摘录中关于折扣增益重视前排相关性的定义、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:Discounted Cumulative Gain (DCG) is the metric that rates and evaluates the gain of each item based on its position in the recommended list of an RS. DCG assumes the earlier arriving items to have more preference. Discounting based on the item position uses a logarithmic factor, as shown below.