Please share good ml interview questions here.
I have the following:
- what are the fundamental reasons why you choose MSE vs MAE loss function?
- What are pros and cons of L1 and L2 regularization.
#Good machine learning interview questions
opaque tartan
rancid spire
tbh I will only ask
What is one machine learning project that you did? (Hopefully the project has some depth) And ask follow up questions all the way (did you use dropout and what is dropout etc, justify your loss function, how would you improve from here)
For knowledge questions, I will only ask these two
- How does a gradient boosted decision tree work
- What is the transformer architecture
I don't see the point of asking random knowledge questions
If the role is domain specific, e.g. recsys - how would you design (company) (product) recsys / what is one issue and how would you improve it?
opaque tartan
One interviewer told me the answer to "reasons why you choose that loss function" is because of the data distribution. For example, in MSE, the square term comes from the MLE estimation of Gaussian distribution.
He want someone who knows why instead of blindly trying different options.