Pinterest measurement tech 电面

miaogemiao 发表于 2019-8-22 01:26
HR 是这样告诉我的，然而我挂在了take-home 上。。。

请问take home是什么题啊？你还有data可以分享下嘛。。。

miaogemiao 发表于 2019-8-21 11:26
HR 是这样告诉我的，然而我挂在了take-home 上。。。

what...楼主好人会有好报的！加油啊！

Frances0412 发表于 2019-8-22 02:21
请问take home是什么题啊？你还有data可以分享下嘛。。。

有点久远了，记不太清了。。。大概是一个估计每个广告商每个月会赞助多少钱，大概有5个variable把，我先做了一下classification， 选取了部分数据， 然后再做的model

variable variance in decision tree那题能不能请教一下，不是minimize variance in allocation也就是target variable吗？那split variable的variance又有什么影响呢？

应该选variance低的呀
Here are the steps to split a decision tree using reduction in variance:
For each split, individually calculate the variance of each child node
Calculate the variance of each split as the weighted average variance of child nodes
Select the split with the lowest variance
Perform steps 1-3 until completely homogeneous nodes are achieved

Myriag 发表于 2020-10-20 08:20
应该选variance低的呀
Here are the steps to split a decision tree using reduction in variance:
For  ...

请问为什么split with the lowest variance 呢？

Given a single decision tree and two variables, variable A has high variance, variable B has low variance, which variable will first be used to split?

感觉这题描述不太清楚。我觉得应该是问: Spliting using variable A gives high variance (其实应该是分割后两个child node的weighted variance), spliting using variable B gives low vairance, which variable should we use to split?

应该选B，就是split之后variance低的。当然是希望split之后的子树越pure越好，pure = low vairance

参考https://youtu.be/i5wGWV5H7oM