Write a summary and give back to the cattle~ For yourself and for the students who will be interviewed in the future, avoid detours, and the hit rate is higher~ I wish everyone a lot of satisfied offers (late summary [Algorithm Post-Machine Learning Direction])
direct to dry goods
Preliminary preparation
Theory: Statistical Learning (Li Hang) + Machine Learning (Zhou Zhihua) + Recommendation System Practice (Xiang Liang) + Deep Learning (That Flower Book) + Introduction to Data Mining + The Beauty of Programming + Sword Offer [Book]
Lin Xuantian of National Taiwan University (I personally think it is very good) [Video]
Niuke.leetcode [website]
Project: You can do kaggle (I feel that this is the real way to improve, the big cows are willing to share technology and can learn a lot) or some domestic competitions (some companies such as Tianchi/datacastle/Meituan Toutiao will also have competitions) Accumulate more valuable self-made core projects
Resume: It can be concise and clear, and the project is the highlight (it is best to have your own main project involving real industrial data, if there is a competition, it will be even better, the icing on the cake)
interview questions
Ali
one side:
Compare the basics, ask projects, algorithms, real-world scenarios
First of all, I will introduce myself. I personally feel that it is better to be concise and clear, and highlight my highlights. For example, I have won the national top several times in other competitions in Tianchi. For example, I was responsible for several projects during my internship. These are the highlights. You can let the interviewer grasp the points you want to ask within 2 minutes, and at the same time give yourself a hot spot (remember, you must be very familiar with the projects or competitions you have done, for example, I am in the direction of data mining, then from data cleaning- >Preprocessing->Business Understanding && Feature Construction->Model Construction This process, as well as the reasons for doing so, the advantages and disadvantages must be known in detail. In short, ask yourself a few more whys. Asking yourself is better than being interviewed The official forced himself to ask dumbfounded.)
Project side:
The first and second interviewers directly asked me to describe the project, but the big guy who crossed all sides did not let me talk about it (will be discussed later);
This is how I introduced my project (competition), first of all, I made clear the problems to be solved by the project, and explained the corresponding application scenarios of the project (for example: this competition is a global competition held by xx, the purpose is to predict experts answering questions) The probability of , that is, whether the expert answers the question and what is the probability of the question being answered, so this is a classification plus regression problem, the application scenario is the recommended question and answer aspect of xxxx, first put the data blabla...) The interviewer will based on your The project description, interrupting you intermittently, interspersed with asking you theories && the advantages and disadvantages of doing so (see clearly that there are advantages and disadvantages, especially the disadvantages), and then extend some basic knowledge, such as XGB/LGB/LR/SVM/ The advantages and disadvantages of three major categories of decision trees/boosting bagging, how to measure the quality of a model (corresponding evaluation indicators auc roc true positive rate, false positive rate, etc.), and a description of the corresponding algorithm process (in the way you understand it to describe ), and sometimes the interviewer will ask about the time complexity and space complexity if they are happy. In short, the advantages and disadvantages of the theory, the scope of application, these are ok, the theory is still relatively easy.
Algorithmic side:
Very simple one. . . 0-1 backpack, binary tree search; string; array (since the interview is the free play of the interviewer, I am afraid that it will limit the interview points of subsequent students, only the scope is given here)
Actual scene questions:
I am referring to the recommendation aspect. The title means that you can use all the data of the Tao Department to design features and models to predict the products that the user is most likely to click (my answer: features, the last 3 days/1 week/ Browsing products in 2 weeks (combined with time series analysis), users add products, users click products, and other behavioral characteristics; there are also user portraits, item portraits, and the corresponding evaluation standard is AUC/LOGLOSS; the model first uses LR to determine whether the corresponding features are valid , after the effective, the common LGB...then go to A/B test)
Second side:
Projects, Algorithms, Practical Scenario Questions
Just finished dinner one night, suddenly a phone call came, and then the interview began. . . . .
project:
Here the interviewer was interested in my academic paper, and then said something (in terms of gene sequencing, because it is not the same as the common one, I will not introduce it)
Then I talked about those two competitions (I picked two competitions to talk about, that is to say, I have prepared about four projects), and still narrated from those aspects.
Then we talked about the project during the internship, and the other party was quite satisfied.
algorithm:
My academic paper involves editing distance, so I asked about this, and bitmapper (this is also my academic paper aspect...)
Actual scene questions:
This is for the internship in Meituan. If I were to increase by 5 percentage points on the original basis, what would I do? (The effect of model fusion is definitely not as good as that of feature fusion and high-level features, so I answered according to this idea, satisfied)
Three sides: hr
I don't know why the three sides are hr. . .
hr asked the basics, asked about the project, and the difficulties encountered by the project, how did I solve it, and whether the data was confidential, etc., and then asked me to compare the offers in my hand and asked me which one I prefer (of course Ali...), and then the other party asked me about my expected salary, and I said what other companies gave and so on.
Four-sided intersection (p9 level)
After a day passed, the interviewer in charge of the second face said on the phone that he wanted me to go further, and made an appointment for me at the cross face, a p9 interviewer (then I asked some cross face cow friends, the butter was very good. Tell me politely and enthusiastically, don't worry, it's just some basic chat and so on, however...)
First ask what offers are there and then direct the project
project:
During the whole process, I did not take the initiative to introduce, but directly asked about the technical points (if you are not in the state or a little unfamiliar, then it will be miserable); directly asked if this project is the data of the real scene, what did you do? (I introduced a little bit, and then was interrupted to ask for details, and then the other party asked how efficient it was. Do you think the efficiency will be higher than if you don't do it? Why?)
Then which models are used (bagging of LGB+XGB) and why? what is the benefit? What parameters did you use in the LGB? How do you adjust the parameters?
Then I saw my academic paper, then I asked questions, and finally finished (in terms of gene sequence, comparison... niche... I talked about my innovations, the other party is not very clear, after all, it is not a field, no for reference)
algorithm:
I asked a few big data related, two files, especially large files A and B, A50G, B10G, one computer, 2G memory, infinite hard disk, let you ask for the intersection and difference of the two files (I use I was a little nervous at first, ignoring that there is only one computer, the boss immediately said that there is only one, and then guided him patiently. Fortunately, I responded quickly, grasped the essence, and answered it. )
Then I asked the question of divide and conquer, and also asked which courses (Introduction to Algorithms, Hadoop, Multi-Core Parallelism...), Boss: Oh, that's a very thick calculation guide, do you choose to study or study the whole book? Who taught it (selection, taught by my boss); boss: Have you studied machine learning? (No, self-taught); Big Brother: Well, do you have any questions for me? (Do you think there is another better way to answer the intersection of those two files?) Big guy: Didn't you answer well, just wanted to see if you had a MapReduce idea; then the big guy asked me Are there any friends in the lab who want to go to Ali? We also have a special concave, let me push him (so kind boss emm, in addition, those who want to come to Ali's 2018 can also stop me, the main recruiting algorithm, social recruiting ( Well-known companies) also recruit)
Meituan
Meituan is three rounds (also a remote phone call, here is the Meituan internship, and finally went to the internship and got the sp)
The first round asked about internship offers, introduced the project, and asked questions about the project
The second round of video interviews, asking about algorithms, asking about basic theoretical knowledge, asking about projects, and asking about codes
In the third round, I asked about the first two rounds, and then focused on the code. I like to ask questions about strings and binary tree search and sorting, followed by projects, and then theory (seeing how important theory is)
The fourth round of hr notification offer
The atmosphere of Meituan is very harmonious, the mentor and the boss are very good, and the problem solving and entry point are also very good. In short, I learned a lot, especially the sentence of the mentor, work for life (mentor is a literary fan) [ Finally got the sp of the algorithm]
iFLYTEK Big Data Research Institute
It is also similar to the three rounds and Meituan, but it is only the scene. The final sp offer
Huawei
I don’t know if it’s to speed up the process or what. I’m just two-faced, one round of technology and one big boss. The technology is talking about my own project, and then I asked a simple algorithm question. The second round of bosses are talking to me about their company. . . . emmmmmmm last sp offer
ZTE
At the end of the four rounds [the question is nothing to say, oh yes, his interview is more interesting, each round is 2 people to meet you one, the pressure side? ? ? But basically no pressure. . . ]
Finally tired, WeChat and Microsoft are pigeons. . .
In addition, I would like to thank my classmates and seniors who have helped me, as well as my good teammates~~~ Knowing you has improved my happiness index~~~
Seeing the students here, I wish you many offers, refill~
Author: Sky_1023
This article comes from Niuke.com
——————————
Niuke.com (www.nowcoder.com)
- Real Questions for Written Exam of Internet Famous Enterprises
- On-campus job-hunting pen & face
- Programmer job placement information
- Programmer learning exchange community