FB DS Onsite面经

northpoint 发表于 2018-10-20 07:40
请问2.2是问percent of receivers who have ever reply

yes~~~~~~~~~~~~~~~

论坛匿名用户 发表于 2018-10-20 09:02
yes~~~~~~~~~~~~~~~

谢谢！希望一切顺利～

follow up: applied data 表现不是很好 。加面一轮：restaurant recommender, 班上有面经。如何建这个推荐系统。用什么数据。怎么验证。每天什么时候发送。如何定义一个受欢迎的餐厅。这个功能和朋友推荐，page推荐有啥不同 。   

2) 2.2) don't forget to group by all three columns in left table, to avoid user_a send message to user_b on different ts.

楼主能讲一下这个是啥意思么？谢谢！！

cathy喵喵喵 发表于 2018-11-6 10:11
2) 2.2) don't forget to group by all three columns in left table, to avoid user_a send message to us ...

consider corner case
user_1, user_2, timestamp
A,B,1
A,B,2
B,A,3
B,C,4
A,B,5

Thanks for sharing

论坛匿名用户 发表于 2018-11-6 05:20
follow up: applied data 表现不是很好 。加面一轮：restaurant recommender, 班上有面经。如何建这个推荐 ...

感谢分享！可以分享下加面的那轮restaurant recommend 和朋友推荐，page推荐有啥不同这问的思路吗？谢谢！

请问楼主，第三道概率题，你的答案是：
P(3_rate_good)=p(3_rate_good|normal_reviewer)*P(normal_review)+p(3_rate_good|lazy)*p(lazy)=0.6^3*0.8+1*0.2
计算量3 posts all reviewed by the same type of reviewers. 我的思路是probability of all 3 posts were rated as good by any type of reviewer可以吗？计算出来是
P(3_rate_good) = [(.8)(.6) + (.2)(1)]^3

同楼上，请问 “restaurant recommend 和朋友推荐/page推荐有啥不同”这个题应该怎么答呢？
谢谢楼主！！

求问解释CI, mean 是怎么回答的呢