Huawei Data Analyst Technical Interview experience
Foreword
Huawei's like the end of the various departments are hiring. I received a invitation to interview, the data analyst. If only three steps, your resume, interview techniques, business interview. It is likely that we most want to know the technical interview and share with you.
text
First, in the form of interview isPhone interviewTime is probably35 minutes( God knows I can talk so ).
The first is the firstGreetingsA few. Asked which city I'm in, then it is a sign Deco emphasized that this is the basis for recruitment. Currently the agency recruit and below 17 are both Deke Ha. As for the relationship Deco and Huawei, we recommend all of you to Baidu. (Of course, if after I signed, I'm sure it will share experience in this area)
Then said technical interview, firstSelf introduction.
I: "XXXX" (about three minutes)
Followed byAsk your resume and project data analysis.
Because my job is to do image recognition and image of the company to do just me. So all steps in this regard I am very clear, and this is the data analysis and can wander relations. So the answer is considered very good.
Then it isAsk questions of data analysis, Guess should be investigated to understand the data analysis. We recommended to understand the next data pre-processing and related algorithms.
Problems are
1. how to measure a model is a good model.
Me: "How do you measure the quality of a model should have five points, accuracy, speed, strong type, the scale, interpretability and then five points I have cited the example of their own."
2. How to optimize a model.
Me: "model of optimization, will depend on what point you are not satisfied with the accuracy of the model if you are not satisfied, you can try the plurality of weak classifiers to form a strong classifier; not satisfied with the speed, you can import numba. then add @jit in front of the function; on the other targeted to change not satisfied enough that a bunch ... XXX "
3. How to improve the accuracy of the model.
I would like to give some experience of others, before the online experience to see some of the great God kaggle share, this time came in handy. I did not eat pork, but I've seen a pig ran. Someone else's favorite sayings, out of my mouth, that's mine.
4. The accuracy of the data is too large to be how to improve.
To be honest, I do not know this, do not know. My answer is, "You mentioned the amount of data one hundred million level, my first time to take into account is the question." I hit a little too close, then said is not clear.
5. The maximum amount of data you're dealing with what level, millions? Tens of millions? One hundred million? One billion?
. . . . . . Sorry, it is up to one hundred thousand (TAT)
6. When you do image processing, how to prevent the impact of data loss during processing of the results?
Speaking of image, the moment I can learn by analogy. Depth study in the pooling layer side-tracked from image pyramid. In the final analysis, it should be is a characteristic of the data is not lost on the line.
7. Do you think you've got a lot of training data, at this time, there has been some new data, do you think these new data is useful to you.
My first answer is to look at whether these data have features I need, then I make a decision. Then I changed my answer, changed helpful. I will put my data to the test set to go.I do not know the answer wrong.
8. Have you ever used the python done some reptilian project?
A. "No, but I did QuickMacro python. Python realize wizard button function, super-simple detail
This is probably eight, no more questions, and then ask them and I said, I had a technical interview, there is still a business side, this interview will not ask questions of the technology. Then I evaluated the breadth enough, deep enough. (In this affirmation, ah, ah my depth image)
The next step is I asked him, I wrote this part of the information I acquired.
Enterprise sector is the delivery and service, Huawei is responsible for the country's AI project. This job base in Chengdu, but need to travel frequently as short as twelve weeks, or as long as six months. Most business trip to Beijing and Guangzhou. Incidentally, the source of the phone interview is in Guangzhou.
Time to go along with the project work (this we know everything)
As for salary, you want to come will be more interested. To be honest, I did not ask, then there is a business side, at that time the interviewer will certainly take the initiative to ask you, first time to exorbitant demands better.
Preparing for the interview
Since the adoption of technical, then share with my preparation, to give you a reference.
1. Understand the technical analyst is doing this to search under their recruitment needs to know, there are also some relevant knowledge.
2. Understand "Ten algorithm data mining" are C4.5, k-Means, SVM, Apriori, EM, PageRank, AdaBoost, kNN, Naive Bayes, and CART. Understand their advantages, disadvantages, applicable data types and scenarios. On the part of the algorithm which is used to run the code a bit, feeling a bit.
3. The knowledge database, catching my mySQL. (Although basic interviewer did not ask)
4. accumulation of normal, I feel quite useless. Daniel say some views and ideas on problem solving, feeling a little used.
Note
This article and my hair before the interviewer to communicate a bit more concerned about privacy issues, but still gained allowed. Tell me what you want to be useful, and trouble can point to a more praise and attention. It promises more weeks. There can not clear message to me, we will promptly reply.
