知识卡片
交替最小二乘按块求解
交替最小二乘固定一组因子,求另一组的条件最小二乘解,再轮换更新。它与把两组参数同时做梯度下降不同,具体求解还受目标和约束影响。迭代减少目标不保证达到全局最佳或带来更好的推荐质量。
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
computes each latent factor by fixing the other factors in two MF models. The next phase is the POI recommendation, where all predicted POIs categories are treated as candidates. Collaborative filtering (CF) is applied with two approaches, namely User-based Collaborative Filtering (UCF) [31] and User-Time-based Collaborative Filtering (UTCF) [31]. UCF deals with a user’s preferences on a particular item and aggregates all of them having
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:摘录中关于交替最小二乘按块求解的定义、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:computes each latent factor by fixing the other factors in two MF models. The next phase is the POI recommendation, where all predicted POIs categories are treated as candidates. Collaborative filtering (CF) is applied with two approaches, namely User-based Collaborative Filtering (UCF) [31] and User-Time-based Collaborative Filtering (UTCF) [31]. UCF deals with a user’s preferences on a particular item and aggregates all of them having