The author | Ivyoake finishing | NewBeeNLP

I have received the letter of intent this morning for the position of Algorithm engineer – Machine learning in Hangzhou. \

One side (resume for half an hour)

I got a phone call while crossing the street and stood on the sidewalk for half an hour.

  • Basic Information (Research direction and internship)
  • Detailed questions were asked about the super-resolution algorithm work during the internship: how to generate data, explain how the network can learn LR and SR mapping from the perspective of probability, how to build and train the network, and how to solve the problem of model landing
  • After understanding the problems encountered by the respondent in overtime-sharing, I made relevant inquiries about the cutting-edge technologies in the industry. What GAN models were used? How to design the Loss function of GAN model and why
  • The interviewer asked why he changed his direction from deep learning to machine learning, and arranged a detailed interview two days later.

Second interview (video interview for 40 minutes)

  • After introducing a machine learning project, I inquired about the project pipeline and related knowledge points
  • Bagging variance reduction and Boosting deviation reduction
  • Introduce the improvement of GBDT by XGB, the improvement of XGB by LGB, and why XGB is used since LGB is used
  • Introduce stacking model integration and why model integration works
  • Programming problem: Find the longest contiguous integer array length in an unordered array. In addition, radix sort and quicksort are examined

Three sides (P9 cross side for one hour)

  • The two periods of internship experience of the main solution, what role to assume respectively, what to do, and simply investigate how to solve the practical problem
  • Detailed inquiry is made on the data processing method in the project, how to use the multiple data sets generated, what problems should be investigated in the processing of missing values, whether the mean filling is scientific, etc
  • The specific features derived from feature engineering are asked in detail, why such features are generated, what is the basis, why PCA is used for dimensionality reduction, and why there are problems when multiple features are highly collinear
  • Why make extensive use of tree models, and what are the advantages
  • How does XGB handle missing values, LGB differential acceleration and the underlying code of the histogram algorithm
  • Open question: On Double 11, coupons are issued to users in the hope of maximizing profits at a certain cost. How to model and issue coupons to users? Users can not do AB tests, how to delineate positive and negative samples?
  • Math problem: for A line segment of length 1, take two points A and B randomly and find the probability density function of length AB

Four sides (40 minutes) :

  • Whether the performance optimization of machine learning algorithm has been done, and the content of superfraction algorithm optimization in the project is introduced
  • Why discretize continuous values, and what are the advantages of doing so
  • What difference does it make if the LR for dichotomization of the last layer of the stacking model is discrete or continuous
  • The generation method and thinking Angle of weak model feature are explained in detail

Five interviews (1 hour and 50 minutes of on-site interview) :

Four evening received the notice that goes to the spot interview, experienced the time since spring recruit and autumn recruit the longest and the most nervous an interview, fortunately the interviewer is very good, eased a few

  • It introduces in detail the personalized recommended projects of the first internship, the division of labor and cooperation within the team, and what fields they are responsible for. How to evaluate and test the generated model scheme? Do you encounter the problem of sample imbalance when processing samples? How to solve it
  • Explain stratified sampling and reservoir sampling respectively
  • For the second internship experience, I asked about GAN model, multi-frame model and loss design
  • The pipeline of machine learning project is introduced in detail. The questions asked are partially repeated and will not be described here
  • Open question: Based on the previous interview record, I would like to continue to ask if you have a deeper understanding and better idea of the coupon issue
  • Programming problem: there is log log, each row has two columns (user ID, commodity ID visited), if two users have access to a commodity ID record, then the relationship index of the two users plus one, find all users in the relationship index of the largest TopK; If all data cannot be stored in the memory, how to solve the problem? Can hadoop be used to solve the problem, principle and idea? Is it possible to slice properly so that sorting results on distributed machines are aggregated into final results?
  • Thinking question: There is A bridge. It takes 25 minutes to cross A, 20 minutes to cross B, 10 minutes to cross C, and 5 minutes to cross D. Only two people can cross A bridge at the same time, and the fast one has to wait for the slow one to arrive. When you walk on the bridge, you must have a flashlight to pass, and there is only one flashlight. How can you make four people pass in 60 minutes

HR side (30 minutes) :

After the on-site interview, the supervisor told me that this round was the last technical interview, and how to have a follow-up interview is the HR interview. On the return subway, I received the telephone interview notice for the next day. I was very happy

  • Introduce yourself
  • Describe your characteristics and what you are good at
  • Tell me what you think of the two companies based on your two internships
  • Talk about your expectations for a company
  • What do you think is your greatest strength at work
  • What direction do you want to be engaged in? Is it scientific research or engineering
  • What are your hobbies
  • Do you have any questions

– END –

Machine Learning Online Manual Deep Learning online Manual AI Basics download (PDF updated to25Set) site QQ group1003271085To join the wechat group, please reply to "add group" to get a discount station knowledge planet coupon, please reply to "knowledge planet" like the article, click on itCopy the code