知识卡片
补全结果是预测而非观测
张量分解利用已有记录学习潜在关系,再对未观测位置给出估计。所谓补全或更稠密张量,是预测分数增加,不是获得了真实到访或喜好标签。验收必须留出真实观测,不能把模型补值再当作事实验证自己。
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
原书出处 · 1 处
- · 查看原书依据
missing values are recovered from the sparse tensor using the outer product of the four output factor matrices to recreate a denser tensor. The new tensor then measures user associations, locations, time slots, and weather. The candidate locations list is accessed for recommendation purposes by selecting N corresponding locations with the highest predicted weights values from the upgraded tensor and ranked accordingly, providing Top-K locations
参考来源
- 位置:《Artificial Intelligence and Data Science in Recommendation System: Current Trends, Technologies, and Applications》第11章《张量分解的兴趣点推荐》(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-C11.html)
- 结论依据:摘录中关于补全结果是预测而非观测的定义、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:missing values are recovered from the sparse tensor using the outer product of the four output factor matrices to recreate a denser tensor. The new tensor then measures user associations, locations, time slots, and weather. The candidate locations list is accessed for recommendation purposes by selecting N corresponding locations with the highest predicted weights values from the upgraded tensor and ranked accordingly, providing Top-K locations