注册一亩三分地论坛,查看更多干货!
您需要 登录 才可以下载或查看附件。没有帐号?注册账号
x
发现几个月不登录会被扣分求加加积分 看不了其他同学的帖子啦
11月内推的Google Research Intern
两轮research 面
第一轮快速过cv, 然后问了两个ml design的问题, 大概就是一个问设计一个sentiment analysis系统,另外一个问设计一个language detection系统
第二轮全部问发过的论文细节
ML design感觉要专门花时间准备准备 有一定的套路,地里面和网上也有
比如这个我就觉得写得很好,主要是好好分析讲清楚
1. Identify the core problem: The interviewer usually ask you to solve a very open-ended design question. But it is usually related to the product that they are working on. Thus, in this step, you need to ask any question you have regarding this problem. What is the goal of this problem? What is this problem trying to solve? Formulate the problem and determine whether this is a regresataset. Are you going to use cross validation? Do you use entropy to evaluate the model or PR curve or RoC curve?
Iteration and optimization. After the initial model and evaluation. If all goes well, you can think about some other ideas for potentially improve the model. For example, how do you handle the cold start problem in recommendation system? Can you come up with more features to help the improvement?
|