2018 Spring Recruitment Internship Interview (Continuous Update)

When I get older, my memory is not good. I will record it here while some things are still in memory. Looking back on the previous interview experience, there were losses and gains. 2017 is the year of the outbreak of artificial intelligence, and major companies staged a battle to grab people, making artificial intelligence students and researchers become the sweet pastry of the market, and many development positions have also begun to transform, and even many other industry personnel also came. This trip to muddy waters. Standing on the tuyere, pigs can climb trees, but in just one year, the tuyere was blocked by pigs, which can also be seen from the 2018 spring recruitment internship. I have to say that the algorithm post still depends on the degree and school. The hardware is not very good, and the resume screening cannot be passed, or even if it passes the screening, it is still in a state of waiting in line for notification. The author delivered AI unicorns such as Megvii Technology and received several rejection letters. I didn't know who I heard before saying that the written test is not important, the key is the interview, so I didn't do much preparation. I did the written test of several companies one after another, but none of them were written, and I almost doubted my life. I have met with several companies and accumulated some experience. I should share it and encourage myself.

Ali Taobao side (graphic image):

1. Self introduction
2. You voted for an image post, tell me about your project experience in imagery
3. What is the purpose of your classification of soybean leaves?
4. Requirements for classification accuracy
5. What is your general approach to deep learning classification?
6. how your image tags are done
7. The role of convolution, pooling, and full connection (to be detailed)
8. Why not use fully connected layers instead of adding convolutional layers
9. What are the methods of pooling? Why use max pooling? Doesn't max pooling lose a lot of information?
10. What activation function is used in the fully connected layer, the difference between sigmod and relu, why not use sigmod in the fully connected layer
11. Application example of radix sort
12. Multi-way merge problem (merge sort of multiple sorted arrays)
13. The complexity of merge sort, whether it can be done without auxiliary array (that is, the space complexity is 1), and the array cannot be modified (that is, the array is read-only)
14. The complexity of heap sort, the meaning of n, the meaning of logn
15. Find the k-th largest number in the array, requiring less complexity
16. Do you understand automatic state machines?
17. any questions to ask

This is the first interview. In fact, the questions are very basic, but the answers are not very good...

CVTE Central Research Institute (Visual Computing):

1. Talk about overfitting
2. Why dropout can prevent overfitting
3. What are the optimization methods
4. The difference between SVM and LR
5. Ways to prevent overfitting
6. Talk about SIFT
7. Which networks have you used
8. DeepID network internal structure
9. Do you understand Newton's law?
10. Eigenvalues ​​and singular values ​​of a matrix
11. The optimization problem may fall into a local optimum, how to find the global optimum solution
12. have any questions to ask

Today's headlines (IOS development, video streaming direction):

1. Self introduction
2. Talk about resume projects (half an hour)
3. What properties are there, briefly introduce them separately
4. GCD operation example
5. The difference between load and initialize
6. Talk about ARC
7. Talk about the Runtime mechanism

Two sides of today's headlines (IOS development, video streaming direction):

1. A little more storyboard for development or pure code
2. Used those tripartite frameworks
3. Talk about KVC and KVO, KVO implementation principle
4. Difference and connection between NSOperationQueue and GCD
5. Understanding of Objective-C (strengths, weaknesses, dynamics, etc.)
6. Let's talk about the proxy model
   ...
   a lot of people forget

VIVO Camera Technology Research Institute (AI Image):

1. Self introduction
2. Talk About Resume Projects
3. Image enhancement and sharpening
4. Improvement of Image LBP Features and Rotation Invariance
5. Does what you do have actual product application?
6. …………I can't remember a lot of them in the middle, but they are relatively basic
7. have any questions to ask

VIVO Camera Technology Research Institute (AI Image) HR side:

1. Self introduction
2. Is there a problem with offsite internships?
3. How long can I train for?
4. Family situation, etc., etc., are more complicated
5. Ask the most concerned question

Huawei side (computer vision):

1. Self introduction
2. Which school is the best in Wuhan? Wuhan University of Technology has a great reputation in recent years, and it is also very good, right?
3. What classification algorithms are there? Briefly talk about
4. SVM kernel function
5. Talk about the first item on your resume
6. Tell me about your experience in the Mathematical Contest in Modeling
7. What are you blogging about? After I finished speaking, I opened the blog and asked a few questions in the blog
8. Is there any consideration for the place of work?

Huawei Two Sides (Computer Vision):

1. Introduce yourself
2. You are from?
3. What are your usual hobbies? (I rap, sing, travel, etc.), and then he actually asked, how was the singing? (Would you like to be ugly?)
4. Have you ever been a class officer? (I said that I have been a deputy monitor for several years in my undergraduate degree), and then he asked again, what does the deputy monitor do? (meaning that an official should be a righteous one)
5. Also asked about digital models and blogs
6. Huawei has a lot of work pressure and overtime. Why come to Huawei?

Meitu Xiuxiu (algorithm):

1. Introduce yourself
2. chat resume project
3. Your project is mainly based on images, do you consider it to be a recommendation algorithm?
4. Chat throughout the whole process, ask which work place to choose (Beijing, Xiamen, Shenzhen)? I said that as long as it's not Beijing, I can accept everything else (I'm not saying that Beijing is not good, it's my personal preference), and then I found out that the interviewer came from Beijing, embarrassing, the position is also from Beijing, embarrassing, I quickly added an internship in Beijing. fine.
5. ask questions