[Interview] @Appier, Machine Learning Scientist
20170714@Appier, Machine Learning Scientist
Interviewer: one RD in ML team
Interview Procedure:
- Self introduction & QA
- student entrepreneurship experience:
- fin-tech domain knowledge
(no any response)
- android development
(no any response again)
- scrum management experience
(no any response again)
- project experience related to deep learning
- online handwriting recognition application in VR
(start to discuss!, QA!)
- traffic flow prediction using LSTM
(discuss)
- SSD algorithm fine-tune project experience/ Deep Q-learing
(...I should not put unfamiliar projects here...)
- Leadership
- kendo sport team leader/ backpacker experience/ international volunteer/ some small competitions/
(no any response again)
- Machine Learning Basics QA
- given you some data, users, favorite songs, historical data...features, how do you recommend new songs to them. what's model/ method will you use to fit/ predict data?
(I really suck at this section, because I have few experiences on basic ML field such as HMM, SVM... etc, maybe it's time to plan learning those basics, hope that I could prepare well next time...)
- Whiteboard Coding
- Given n classes please implement one function that return random selected classes id with distribution w as array .
(very simple question to verify your programming skills)
Conclusion:
- Basically, the interviewer only focus on the technical part of my experience, which depress me a lot. seems like the company only appreciate people who is best in specific related field, and don't accept differences.
- The whole interview experience: HR is so kind, interviewer also not bad but the questions he asked are felt like free style, ask any questions came out to him (or maybe just I really suck at ML basics concept), which make me very nervous during interview....
- Appier is an appealing company, not sure if I have the opportunity to join them. waiting their rely.
LSTM stands for Long Short-Term Memory, a type of recurrent neural network (RNN) designed to learn patterns in sequential data while overcoming the limitations of traditional RNNs. It uses specialized memory cells and gates to retain important information over long periods, making it highly effective for tasks such as language translation, speech recognition, text prediction, sentiment analysis, and time-series forecasting. Because of its ability to capture long-term dependencies, LSTM remains one of the most widely used deep learning models for sequence-based applications.
ReplyDelete