本帖最后由 Chasedream.df 于 2022-4-28 19:47 编辑
这两天有群友问为啥我的实战贴里面有一些是机器学习pipeline的,对一般swe是不是需要掌握,答案是不用的,但我背景比较特别,在大厂的AI组下面做后端开发,也跟很多mle ds打交道,看他们怎么模型搭建,evaluate,我对这一块就开始自学,在coursera和google公开课上都可以找到很多,所以在求职上就让很多公司觉得我背景也适合ML infra,经常问的是推荐系统设计,比如feed推荐,视频推荐,购物推荐,正好准备了这个就拿下,但说实话这仅仅是机器学习的冰山一角,要是追深不同领域的话,就不好说了。如何提升是要花很大精力的,数学基础,经典paper,公司博客那么多(需要学的我在后面都会列出了),并且你要跟最聪明的人一起竞争,说实话要是为了一个job机会这ROI太低了。
不同公司的叫法,mle,ai infra,learning dev. From 1point 3acres bbs
.
来自google developer社区的推荐系统 https://developers.google.com/ma ... tion/overview/types 很推荐入门. From 1point 3acres bbs
下面来两道原题,各位大神有好的思路也欢迎留言。
snap问的
You're joining the ads ranking team. You find out that the ad delivery system is optimizing for the company and advertisers.
Change the system to balance more value towards the users as well.
. 1point3acres.com
另一个fintech公司问的
Problem Statement:
.google и
* Improve Average Order Value. i.e increase the $$ spent by shoppers on the site.
.1point3acres
* Idea: Build an API to show:
- Complimentary products (things that were purchased together)
. check 1point3acres for more. - Input: product_id
- expected output; complimentary_products[]
- Alternatives (say you are looking at a phone case. Suggest other brands and alternatives).google и
- Input: product_id.google и
- expected output; similar_products[]. .и
Existing state of the art:. Waral dи,
- We have an orders tables:
- order_id.--
- cart_id
- txn_id
- user_id. Waral dи,
. From 1point 3acres bbs
carts: (many to many)
- cart_id <> product Id mapping.
txn:. 1point 3acres
- txn_id,
- cost
- taxes
CC /Paypal etc.
users:
- user_id-baidu 1point3acres
- email
- phone
- address. .и
-baidu 1point3acres
这里就把meta跟google的准备资料给大家. check 1point3acres for more.
. 1point3acres
. 1point 3 acres
. check 1point3acres for more.
|